fix: stop declaring max output tokens (#1479)

Author: edwinyjlim

src/agent/runner/harness/pi/gateway.ts

@@ -97,7 +97,8 @@ export function buildGatewayModel(inputs: GatewayProviderInputs) {
     // UnsupportedParamsError → the run no-ops).
     reasoning: modelCapabilities(modelId).reasoning,
     // The gateway refuses `developer` messages on pinned tokens, so pi sends `system`.
-    compat: { supportsDeveloperRole: false },
+    // It also sizes a scoped token's hold from max_output_tokens, so omit it, as pi 0.79 did.
+    compat: { supportsDeveloperRole: false, supportsMaxOutputTokens: false },
     input: ['text' as const],
     cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
     contextWindow: 1_000_000,

src/agent/triage-provider.ts

@@ -58,14 +58,19 @@ export function createTriageLLMProvider(
     const { completeSimple } = await import('@earendil-works/pi-ai/compat');
     const current = await resolveAuth();
     const authToken = current.authToken;
-    const model = buildGatewayModel({
+    const gatewayModel = buildGatewayModel({
       gatewayUrl: current.baseURL,
       accessToken: authToken,
       teamId: current.teamId,
       wizardMetadata: current.wizardMetadata ?? {},
       wizardFlags: current.wizardFlags ?? {},
       modelId,
     });
+    // Triage sets its own small cap, so keep sending it on the OpenAI path.
+    const model = {
+      ...gatewayModel,
+      compat: { ...gatewayModel.compat, supportsMaxOutputTokens: true },
+    };
     const controller = new AbortController();
     const timer = setTimeout(() => controller.abort(), TRIAGE_TIMEOUT_MS);
     try {