fix(agents): resolve context window for models with slash in ID (e.g. OpenRouter proxy models)

samson1357924 · samson1357924 · commit b79687dcfe7a · 2026-06-13T22:44:57.000+08:00
Remove the `!ref.model.includes("/")` guard in `resolveContextTokensForModel()` which blocked qualified cache key lookup for models like `anthropic/claude-sonnet-4` accessed through proxy providers.

Two call sites fixed:
1. Explicit provider path (params.provider &amp;&amp; ref) -- safe because provider is explicitly given
2. Last resort implicit provider path (!params.provider &amp;&amp; ref) -- bare key already failed, trying qualified key has no downside

Fixes: max context showing DEFAULT_CONTEXT_TOKENS (200K) instead of actual model's context window for proxy models
diff --git a/src/agents/context.ts b/src/agents/context.ts
@@ -389,7 +389,7 @@ export function resolveContextTokensForModel(params: {
     }
     // Only do the config direct scan when the caller explicitly passed a
     // provider. When provider is inferred from a slash in the model string
-    // (e.g. "google/gemini-2.5-pro" → ref.provider = "google"), the model ID
+    // (e.g. "google/gemini-2.5-pro" ??ref.provider = "google"), the model ID
     // may belong to a DIFFERENT provider (e.g. an OpenRouter session). Scanning
     // cfg.models.providers.google in that case would return Google's configured
     // window and misreport context limits for the OpenRouter session.
@@ -409,16 +409,16 @@ export function resolveContextTokensForModel(params: {
   // When provider is explicitly given and the model ID is bare (no slash),
   // try the provider-qualified cache key BEFORE the bare key.  Discovery
   // entries are stored under qualified IDs (e.g. "google-gemini-cli/
-  // gemini-3.1-pro-preview → 1M"), while the bare key may hold a cross-
+  // gemini-3.1-pro-preview ??1M"), while the bare key may hold a cross-
   // provider minimum (128k).  Returning the qualified entry gives the correct
   // provider-specific window for /status and session context-token persistence.
   //
   // Guard: only when params.provider is explicit (not inferred from a slash in
   // the model string). For model-only callers (e.g. status.ts log-usage
   // fallback with model="google/gemini-2.5-pro"), the inferred provider would
   // construct "google/gemini-2.5-pro" as the qualified key which accidentally
-  // matches OpenRouter's raw discovery entry — the bare lookup is correct there.
-  if (params.provider && ref && !ref.model.includes("/")) {
+  // matches OpenRouter's raw discovery entry ??the bare lookup is correct there.
+  if (params.provider && ref) {
     const qualifiedResult = lookupContextTokens(
       `${normalizeProviderId(ref.provider)}/${ref.model}`,
       {
@@ -441,10 +441,12 @@ export function resolveContextTokensForModel(params: {
     return bareResult;
   }
 
-  // When provider is implicit, try qualified as a last resort so inferred
-  // provider/model pairs (e.g. model="google-gemini-cli/gemini-3.1-pro")
-  // still find discovery entries stored under that qualified ID.
-  if (!params.provider && ref && !ref.model.includes("/")) {
+  // Last resort: try qualified key even when ref.model contains a slash.
+  // The bare lookup already failed, so there is no downside to trying the
+  // full qualified key. This covers models like "anthropic/claude-sonnet-4"
+  // accessed through a proxy provider whose discovery cache stores the
+  // full qualified ID.
+  if (!params.provider && ref) {
     const qualifiedResult = lookupContextTokens(
       `${normalizeProviderId(ref.provider)}/${ref.model}`,
       {
