fix(openrouter): use endpoint context limits

TurboTheTurtle · TurboTheTurtle · commit f4ca87acaa9a · 2026-05-23T23:16:05.000-07:00
diff --git a/src/agents/pi-embedded-runner/openrouter-model-capabilities.test.ts b/src/agents/pi-embedded-runner/openrouter-model-capabilities.test.ts
@@ -92,6 +92,48 @@ describe("openrouter-model-capabilities", () => {
     });
   });
 
+  it("uses endpoint-specific OpenRouter context length when top_provider reports one", async () => {
+    await withOpenRouterStateDir(async () => {
+      vi.stubGlobal(
+        "fetch",
+        vi.fn(
+          async () =>
+            new Response(
+              JSON.stringify({
+                data: [
+                  {
+                    id: "nvidia/nemotron-3-super-120b-a12b:free",
+                    name: "Nemotron 3 Super 120B Free",
+                    architecture: { modality: "text->text" },
+                    context_length: 1_000_000,
+                    top_provider: {
+                      context_length: 262_144,
+                      max_completion_tokens: 262_144,
+                    },
+                    pricing: { prompt: "0", completion: "0" },
+                  },
+                ],
+              }),
+              {
+                status: 200,
+                headers: { "content-type": "application/json" },
+              },
+            ),
+        ),
+      );
+
+      const module = await importOpenRouterModelCapabilities("top-provider-context-length");
+      await module.loadOpenRouterModelCapabilities("nvidia/nemotron-3-super-120b-a12b:free");
+
+      expect(
+        module.getOpenRouterModelCapabilities("nvidia/nemotron-3-super-120b-a12b:free"),
+      ).toMatchObject({
+        contextWindow: 262_144,
+        maxTokens: 262_144,
+      });
+    });
+  });
+
   it("preserves explicit OpenRouter tool support metadata", async () => {
     await withOpenRouterStateDir(async () => {
       vi.stubGlobal(
diff --git a/src/agents/pi-embedded-runner/openrouter-model-capabilities.ts b/src/agents/pi-embedded-runner/openrouter-model-capabilities.ts
@@ -49,6 +49,7 @@ interface OpenRouterApiModel {
   max_completion_tokens?: number;
   max_output_tokens?: number;
   top_provider?: {
+    context_length?: number;
     max_completion_tokens?: number;
   };
   pricing?: {
@@ -174,7 +175,7 @@ function parseModel(model: OpenRouterApiModel): OpenRouterModelCapabilities {
     input,
     reasoning: supportedParameters?.includes("reasoning") ?? false,
     ...(supportedParameters ? { supportsTools: supportedParameters.includes("tools") } : {}),
-    contextWindow: model.context_length || 128_000,
+    contextWindow: model.top_provider?.context_length ?? model.context_length ?? 128_000,
     maxTokens:
       model.top_provider?.max_completion_tokens ??
       model.max_completion_tokens ??
