Shorten the comments on the newest round of changes

Comment-only pass over the quant-suffix split, the storage-shape normalization
and the wall-clock deadline in the idle-unload test. Same points, fewer lines.

studio/backend/tests/test_openai_auto_switch.py

@@ -566,11 +566,8 @@ def test_idle_loop_deletes_saved_kv_when_unload_fails(monkeypatch, tmp_path):
 
     async def _drive():
         task = asyncio.create_task(kw.idle_unload_loop(poll_seconds = 0.01))
-        # A wall-clock deadline, not an iteration count: Windows timers round a
-        # 10 ms sleep up to the ~15.6 ms scheduler tick, on both this loop and the
-        # one under test, so a fixed count of short sleeps gave the unload far
-        # less real time there than the number suggests and the test failed for
-        # being slow rather than wrong.
+        # Wall clock, not an iteration count: Windows rounds a 10 ms sleep up to the
+        # ~15.6 ms tick on both loops, so a fixed count failed for being slow.
         deadline = time.monotonic() + 15.0
         while time.monotonic() < deadline:
             await asyncio.sleep(0.01)

studio/frontend/src/features/model-picker/api/migrate-model-overrides.ts

@@ -44,13 +44,10 @@ function markRan(): void {
 /**
  * A server key under the same identity this browser stores.
  *
- * `app_settings` has no schema version and holds whatever id was current when the
- * row was written, so an old install has keys like `Unsloth/Repo-GGUF:Q4_K_M` while
- * this browser only produces the folded form. The backend resolves both to one
- * model, so an exact lookup would report "not on the server" and let the backfill
- * overwrite it. The split is quant-aware like the backend's: a repo id folds and a
- * POSIX path deliberately does not, and an ordinary colon inside a filename, or a
- * Windows drive letter, is not a separator at all.
+ * `app_settings` has no schema version, so an old install holds keys like
+ * `Unsloth/Repo-GGUF:Q4_K_M` that the backend resolves to the same model as this
+ * browser's folded form; an exact lookup would call it missing and let the backfill
+ * overwrite it. The quant-aware split folds repo ids and leaves POSIX paths alone.
  */
 function normalizedOverrideKey(key: string): string {
   const split = splitQuantSuffix(key);

studio/frontend/src/features/model-picker/components/model-config-page.tsx

@@ -872,12 +872,9 @@ export function ModelConfigPage({
     const effectiveAtBaseline = perModelConfigsEqual(effectiveConfig, baseline);
     const effectivePersistenceOnly =
       isActiveModel && effectiveAtBaseline && rememberChanged;
-    // Storage's own shape, because savePerModelConfig judges the normalized
-    // object: the runtime hands this page Speculative Decoding "auto", which
-    // canonicalizes to null, so the raw one looked non-default. Remember then
-    // reported saved while the local write had dropped the entry as default, and
-    // the mirror below sent the server a "auto" override the browser did not
-    // have, which is exactly the disagreement that mirror must not create.
+    // Judge what storage keeps: savePerModelConfig normalizes first, and the runtime's
+    // Speculative Decoding "auto" canonicalizes to null, so judging the raw object
+    // reported saved while the write dropped it, and mirrored an override nothing held.
     const normalizedRuntimeConfig = normalizePerModelConfig(
       effectiveRuntimeConfig,
     );

studio/frontend/src/features/model-picker/model-config/model-identity.ts

@@ -69,8 +69,7 @@ export function ggufVariantFromStorageKey(key: string): string | null {
 }
 
 // Mirrors split_quant_suffix in studio/backend/utils/openai_auto_switch_settings.py.
-// A quant label may carry a bits-per-weight modifier ("IQ4_XS-3.53bpw"), and the
-// two backend label helpers disagree on whether to keep it, so both forms parse.
+// The bpw modifier ("IQ4_XS-3.53bpw") is optional: the backend label helpers disagree.
 const BPW_SUFFIX = /-[0-9]+(?:\.[0-9]+)?bpw$/i;
 const KNOWN_QUANT =
   /^(UD-)?(MXFP[0-9]+(?:_[A-Z0-9]+)*|IQ[0-9]+_[A-Z]+(?:_[A-Z0-9]+)?|TQ[0-9]+_[0-9]+|Q[0-9]+_K_[A-Z]+|Q[0-9]+_[0-9]+|Q[0-9]+_K|BF16|F16|F32)$/i;
@@ -79,9 +78,8 @@ const MAX_QUANT_SUFFIX_LEN = 64;
 /**
  * `[head, quant]` for a `head:QUANT` key, or null when the colon is not one.
  *
- * The suffix has to look like a real quant, so an ordinary colon inside a POSIX
- * filename is left alone ("/models/foo:bar.gguf" is one valid filename) and a
- * Windows drive letter is never mistaken for a model id.
+ * The suffix must look like a real quant, so an ordinary colon in a POSIX filename
+ * ("/models/foo:bar.gguf") and a Windows drive letter are left alone.
  */
 export function splitQuantSuffix(value: string): [string, string] | null {
   const separator = value.lastIndexOf(":");
@@ -99,9 +97,7 @@ export function splitQuantSuffix(value: string): [string, string] | null {
   ) {
     return [head, tail];
   }
-  // A .gguf with no recognizable quant token is labelled by its stem, so keys
-  // like "/models/CustomModel.gguf:custommodel" exist. Requiring the head to be
-  // a .gguf keeps an ordinary colon out: "/models/foo:bar.gguf" splits to a head
-  // that is not one.
+  // A .gguf with no recognizable quant is labelled by its stem, so
+  // "/models/CustomModel.gguf:custommodel" exists; a non-.gguf head is a plain colon.
   return head.toLowerCase().endsWith(".gguf") ? [head, tail] : null;
 }

studio/frontend/src/features/model-picker/model-config/per-model-config.ts

@@ -480,12 +480,9 @@ function normalizeV1(partial: RawConfig): PerModelConfig {
 }
 
 /**
- * A config in the exact shape storage keeps it in.
- *
- * Callers that decide anything from a config a user is still editing have to run
- * it through this first: the UI carries sentinels storage does not, notably
- * Speculative Decoding "auto", which canonicalizes to null. Judging the raw
- * object calls a default config non-default.
+ * A config in the exact shape storage keeps it in: the UI carries sentinels storage
+ * does not, notably Speculative Decoding "auto" which canonicalizes to null, so a
+ * config still being edited reads as non-default when it is not.
  */
 export function normalizePerModelConfig(raw: unknown): PerModelConfig {
   return normalize(raw);

tests/studio/test_model_picker_contracts.py

@@ -697,8 +697,7 @@ def test_backfill_compares_server_keys_by_normalized_identity():
     # Folded on both sides: the older `id::variant` local keys are not.
     assert "const known = new Set(Object.keys(existing).map(normalizedOverrideKey));" in src
     assert "if (known.has(key)) { continue; }" in src
-    # A variant never holds a colon, so the last one splits the key; the first would
-    # cut the drive letter off every Windows path id.
+    # A quant-aware split, so a Windows drive letter is not read as a separator.
     assert "const split = splitQuantSuffix(key);" in src
     # Repo ids fold and POSIX paths do not, which is what these do.
     assert "normalizeModelIdentity(" in src and "normalizeGgufVariantIdentity(" in src
@@ -895,8 +894,7 @@ def test_backfill_splits_a_quant_suffix_the_way_the_backend_does():
     ).read_text(encoding = "utf-8")
     assert "def split_quant_suffix(" in backend, "the rule this mirrors"
     assert "_BPW_SUFFIX" in backend and "bpw" in identity
-    # Both sides accept the same quant vocabulary. The regex itself lives with
-    # the loader that reads the filenames.
+    # Both sides accept the same quant vocabulary; the regex lives with the loader.
     quants = (WORKDIR / "studio" / "backend" / "core" / "inference" / "llama_cpp.py").read_text(
         encoding = "utf-8"
     )