From 361854189a7530b673bfb174f81be79a09c0a899 Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 12 Jun 2026 13:54:09 +0000
Subject: [PATCH 1/2] DiffusionGemma: disable tools, enable artifacts canvas by
 default

DiffusionGemma serves via the visual runner, which streams per-step
canvas frames so the answer resolves live in the bubble. The agentic
tool loop (generate_chat_completion_with_tools) does not forward those
frames, so whenever a tool pill (Search/Code) was on the live canvas
silently vanished while text still streamed. DiffusionGemma is not a
tool-calling target anyway, so report supports_tools=False for it: the
chat always takes the frame-forwarding path, and the Search/Code pills
disable themselves (a local model has no builtin web search either).

Also turn the artifacts canvas on by default for DiffusionGemma so a
full-HTML answer (e.g. a playable game) renders as an interactive
sandboxed card without the user flipping the global artifacts toggle.
---
 studio/backend/core/inference/llama_cpp.py               | 4 ++++
 .../src/components/assistant-ui/markdown-text.tsx        | 9 ++++++---
 2 files changed, 10 insertions(+), 3 deletions(-)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index b5d5328fb0..f30a8acf89 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -1042,6 +1042,10 @@ class LlamaCppBackend:
 
     @property
     def supports_tools(self) -> bool:
+        # DiffusionGemma serves via the visual runner, whose live per-step canvas
+        # frames are dropped by the agentic tool loop; never route it through tools.
+        if self._is_diffusion:
+            return False
         return self._supports_tools
 
     @property
diff --git a/studio/frontend/src/components/assistant-ui/markdown-text.tsx b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
index 45a0082a57..61b97e619f 100644
--- a/studio/frontend/src/components/assistant-ui/markdown-text.tsx
+++ b/studio/frontend/src/components/assistant-ui/markdown-text.tsx
@@ -285,11 +285,14 @@ function CodeBlockActions({
 }
 
 // DiffusionGemma renders its denoising live in the bubble (see DiffusionCanvas in
-// thread.tsx), so it no longer forces HTML into an iframe artifact; it follows the
-// same artifact rules as every other model.
+// thread.tsx) and has the artifacts canvas on by default, so a full-HTML answer
+// (e.g. a playable game) renders as an interactive card without the global toggle.
 function StreamdownBlock(props: BlockProps) {
   const shouldCollapseHtmlArtifacts = useChatRuntimeStore(
-    (state) => state.artifactsEnabled || state.collapseHtmlArtifacts,
+    (state) =>
+      state.artifactsEnabled ||
+      state.collapseHtmlArtifacts ||
+      state.loadedIsDiffusion,
   );
   const messageHasRenderableRenderHtmlTool = useAuiState(({ message }) =>
     message.parts.some(isRenderableRenderHtmlToolPart),

From 1c58c00b54bdafbd98d2899bca86f03304a0a9ed Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Fri, 12 Jun 2026 14:17:42 +0000
Subject: [PATCH 2/2] DiffusionGemma: set UNSLOTH_IS_PRESENT for the shim
 subprocess

A clean install could not run DiffusionGemma: the runner spawns
python -m unsloth_zoo.diffusion_studio.shim, and unsloth_zoo refuses to
import unless UNSLOTH_IS_PRESENT is set (normally by import unsloth). The
shim never imports unsloth, so the subprocess died with
'Please install Unsloth via pip install unsloth!' and the model load
failed with a 500. Set the flag in the runner child env, as unsloth does
on import.
---
 studio/backend/core/inference/llama_cpp.py | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/studio/backend/core/inference/llama_cpp.py b/studio/backend/core/inference/llama_cpp.py
index f30a8acf89..a25ed0889a 100644
--- a/studio/backend/core/inference/llama_cpp.py
+++ b/studio/backend/core/inference/llama_cpp.py
@@ -2527,6 +2527,10 @@ class LlamaCppBackend:
         ]
 
         env = child_env_without_native_path_secret()
+        # `python -m unsloth_zoo.diffusion_studio.shim` imports unsloth_zoo, which
+        # refuses to load unless UNSLOTH_IS_PRESENT is set (normally by `import
+        # unsloth`). The shim never imports unsloth, so set it here as unsloth does.
+        env["UNSLOTH_IS_PRESENT"] = "1"
         env["DG_VISUAL_BIN"] = visual_bin
         env["DG_GPU"] = gpu
         # The file-override shim imports its sibling visual_engine; put its dir on PYTHONPATH.