fix(agents): forward model maxTokens by default

steipete · steipete · commit 399d7f61783f · 2026-05-03T14:30:16.000+01:00
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -23,6 +23,7 @@ Docs: https://docs.openclaw.ai
 
 ### Fixes
 
+- Agents/models: forward model `maxTokens` as the default output-token limit for OpenAI-compatible Responses and Completions transports when no runtime override is provided, preventing provider defaults from silently truncating larger outputs. (#76645) Thanks @joeyfrasier.
 - Gateway/update: run `doctor --non-interactive --fix` after Control UI global package updates before reporting success, so legacy config is migrated before the gateway restart. Thanks @stevenchouai.
 - Gateway/cron: stop a lazy cron startup that loses a hot-reload race, preventing the old cron service from starting after reload has already replaced cron state.
 - CLI/plugins: warn when npm plugin installs remain shadowed by a failing config-selected source and surface the repair path in `plugins doctor`. Thanks @LindalyX-Lee.
diff --git a/src/agents/openai-transport-stream.test.ts b/src/agents/openai-transport-stream.test.ts
@@ -995,6 +995,31 @@ describe("openai transport stream", () => {
     expect(params.input?.[0]).toMatchObject({ role: "developer" });
   });
 
+  it("uses model maxTokens for Responses params when runtime maxTokens is omitted", () => {
+    const params = buildOpenAIResponsesParams(
+      {
+        id: "gpt-5.4",
+        name: "GPT-5.4",
+        api: "openai-responses",
+        provider: "openai",
+        baseUrl: "https://api.openai.com/v1",
+        reasoning: true,
+        input: ["text"],
+        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+        contextWindow: 200000,
+        maxTokens: 65_536,
+      } satisfies Model<"openai-responses">,
+      {
+        systemPrompt: "system",
+        messages: [],
+        tools: [],
+      } as never,
+      undefined,
+    ) as { max_output_tokens?: unknown };
+
+    expect(params.max_output_tokens).toBe(65_536);
+  });
+
   it("uses top-level instructions for Codex responses and strips unsupported ChatGPT params", () => {
     const params = buildOpenAIResponsesParams(
       {
@@ -2373,6 +2398,57 @@ describe("openai transport stream", () => {
     expect(params).not.toHaveProperty("max_completion_tokens");
   });
 
+  it("uses model maxTokens for OpenAI completions params when runtime maxTokens is omitted", () => {
+    const params = buildOpenAICompletionsParams(
+      {
+        id: "gpt-5.4",
+        name: "GPT-5.4",
+        api: "openai-completions",
+        provider: "openai",
+        baseUrl: "https://api.openai.com/v1",
+        reasoning: true,
+        input: ["text"],
+        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+        contextWindow: 200000,
+        maxTokens: 65_536,
+      } satisfies Model<"openai-completions">,
+      {
+        systemPrompt: "system",
+        messages: [],
+        tools: [],
+      } as never,
+      undefined,
+    );
+
+    expect(params.max_completion_tokens).toBe(65_536);
+    expect(params).not.toHaveProperty("max_tokens");
+  });
+
+  it("uses model maxTokens with max_tokens completions compat when runtime maxTokens is omitted", () => {
+    const params = buildOpenAICompletionsParams(
+      {
+        id: "zai-org/GLM-4.7-TEE",
+        name: "GLM 4.7 TEE",
+        api: "openai-completions",
+        provider: "chutes",
+        reasoning: true,
+        input: ["text"],
+        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+        contextWindow: 200000,
+        maxTokens: 65_536,
+      } as never,
+      {
+        systemPrompt: "system",
+        messages: [],
+        tools: [],
+      } as never,
+      undefined,
+    );
+
+    expect(params.max_tokens).toBe(65_536);
+    expect(params).not.toHaveProperty("max_completion_tokens");
+  });
+
   it("omits strict tool shaping for Z.ai default-route completions providers", () => {
     const params = buildOpenAICompletionsParams(
       {
diff --git a/src/agents/openai-transport-stream.ts b/src/agents/openai-transport-stream.ts
@@ -981,8 +981,9 @@ export function buildOpenAIResponsesParams(
     ...(isCodexResponses ? { instructions: buildOpenAICodexResponsesInstructions(context) } : {}),
     ...(metadata ? { metadata } : {}),
   };
-  if (options?.maxTokens) {
-    params.max_output_tokens = options.maxTokens;
+  const effectiveMaxTokens = options?.maxTokens || model.maxTokens;
+  if (effectiveMaxTokens) {
+    params.max_output_tokens = effectiveMaxTokens;
   }
   if (options?.temperature !== undefined) {
     params.temperature = options.temperature;
@@ -1863,11 +1864,14 @@ export function buildOpenAICompletionsParams(
   if (compat.supportsPromptCacheKey && cacheRetention !== "none" && options?.sessionId) {
     params.prompt_cache_key = options.sessionId;
   }
-  if (options?.maxTokens) {
-    if (compat.maxTokensField === "max_tokens") {
-      params.max_tokens = options.maxTokens;
-    } else {
-      params.max_completion_tokens = options.maxTokens;
+  {
+    const effectiveMaxTokens = options?.maxTokens || model.maxTokens;
+    if (effectiveMaxTokens) {
+      if (compat.maxTokensField === "max_tokens") {
+        params.max_tokens = effectiveMaxTokens;
+      } else {
+        params.max_completion_tokens = effectiveMaxTokens;
+      }
     }
   }
   if (options?.temperature !== undefined) {
