fix(model-metadata): skip OpenRouter for known providers, add kimi/moonshot to PROVIDER_TO_MODELS_DEV

Based on PR #23950 by @nicoechaniz.

- Add "kimi" and "moonshot" to PROVIDER_TO_MODELS_DEV → kimi-for-coding
- Gate OpenRouter metadata step behind "if not effective_provider":
  known providers should not be overridden by community-maintained OR data
- Keep the targeted Kimi-family 32k guard as a secondary safety net
  inside the OR gate (for unknown providers with Kimi models)

Co-authored-by: nicoechaniz <nicoechaniz@altermundi.net>

agent/model_metadata.py

@@ -1566,27 +1566,33 @@ def get_model_context_length(
         if ctx:
             return ctx
 
-    # 6. OpenRouter live API metadata (provider-unaware fallback)
+    # 6. OpenRouter live API metadata — provider-unaware fallback.
-    metadata = fetch_model_metadata()
+    # Only consulted when the provider is unknown (no effective_provider),
-    if model in metadata:
+    # because OpenRouter data is community-maintained and can be incorrect
-        or_ctx = metadata[model].get("context_length", DEFAULT_FALLBACK_CONTEXT)
+    # for models that belong to known providers with curated defaults.
-        # Guard against stale OpenRouter metadata for Kimi-family models.
+    if not effective_provider:
-        # OpenRouter reports 32768 for moonshotai/kimi-k2.6, but the model
+        metadata = fetch_model_metadata()
-        # actually supports 262144 (models.dev + official Kimi docs agree).
+        if model in metadata:
-        # Providers that host their own Kimi endpoints (Ollama Cloud, Kimi
+            or_ctx = metadata[model].get("context_length", DEFAULT_FALLBACK_CONTEXT)
-        # Coding, Moonshot) would otherwise trip the 64k minimum-context
+            # Guard against stale OpenRouter metadata for Kimi-family models.
-        # guard and reject a perfectly capable model.
+            # OpenRouter reports 32768 for moonshotai/kimi-k2.6, but the model
-        # The filter is narrow: only reject exactly 32768 for Kimi-named
+            # actually supports 262144 (models.dev + official Kimi docs agree).
-        # models.  If OpenRouter ever updates its data, the stale path
+            # Providers that host their own Kimi endpoints (Ollama Cloud, Kimi
-        # becomes dead code with no impact.
+            # Coding, Moonshot) would otherwise trip the 64k minimum-context
-        if or_ctx == 32768 and _model_name_suggests_kimi(model):
+            # guard and reject a perfectly capable model.
-            logger.info(
+            # The filter is narrow: only reject exactly 32768 for Kimi-named
-                "Rejecting OpenRouter metadata context=%s for %r "
+            # models.  If OpenRouter ever updates its data, the stale path
-                "(Kimi-family underreport); falling through to hardcoded defaults",
+            # becomes dead code with no impact.
-                or_ctx, model,
+            if or_ctx == 32768 and _model_name_suggests_kimi(model):
-            )
+                logger.info(
-        else:
+                    "Rejecting OpenRouter metadata context=%s for %r "
-            return or_ctx
+                    "(Kimi-family underreport); falling through to hardcoded defaults",
+                    or_ctx, model,
+                )
+            else:
+                return or_ctx
+
+    # 7. (reserved)
 
     # 8. Hardcoded defaults (fuzzy match — longest key first for specificity)
     # Only check `default_model in model` (is the key a substring of the input).

agent/models_dev.py

@@ -145,7 +145,9 @@ PROVIDER_TO_MODELS_DEV: Dict[str, str] = {
     "openai": "openai",
     "openai-codex": "openai",
     "zai": "zai",
+    "kimi": "kimi-for-coding",
     "kimi-coding": "kimi-for-coding",
+    "moonshot": "kimi-for-coding",
     "stepfun": "stepfun",
     "kimi-coding-cn": "kimi-for-coding",
     "minimax": "minimax",