fix(agent-model): omit synthesized maxTokens fallback when nothing was configured (#98312)

* fix(agent-model): omit synthesized maxTokens fallback (Fixes #98295)

The configured-fallback resolver used DEFAULT_CONTEXT_TOKENS (200_000) as
the last-resort value for model.maxTokens when no configured, provider,
or bundled-catalog value was known. For strict OpenAI-compatible providers,
model.maxTokens is forwarded as max_completion_tokens on every request,
so the synthesized 200k value became a wire-level output cap that exceeded
the provider ceiling and produced HTTP 400 (Param Incorrect) on providers
like Xiaomi MiMo (131_072 max) — reported for mimo-v2.5 / mimo-v2.5-pro
added via 'models add' on a custom provider entry.

The sibling applyConfiguredProviderOverrides path already omits maxTokens
when nothing resolves; align resolveConfiguredFallbackModel with that
behavior. When the transport sees an undefined model.maxTokens, it omits
max_completion_tokens entirely and the provider applies its own default,
which is the correct behavior when the user has configured no output limit.

contextWindow retains DEFAULT_CONTEXT_TOKENS as its local budgeting
fallback (it is not shipped on the wire).

* test(agents): prove unknown output cap omission

---------

Co-authored-by: Peter Steinberger <steipete@gmail.com>

src/agents/embedded-agent-runner/model.test.ts

@@ -1243,6 +1243,39 @@ describe("resolveModel", () => {
     });
   });
 
+  it("leaves maxTokens undefined when no configured or catalog value is available (regression: #98295)", () => {
+    // Regression for https://github.com/openclaw/openclaw/issues/98295.
+    // A custom provider entry without maxTokens (and no matching bundled
+    // static catalog row) must not synthesize an oversized output cap from
+    // DEFAULT_CONTEXT_TOKENS. Leaving maxTokens undefined lets the transport
+    // omit `max_completion_tokens` so the provider applies its own default,
+    // avoiding HTTP 400 (Param Incorrect) from strict OpenAI-compatible
+    // servers whose completion-token ceiling is below the synthesized value.
+    resolveBundledStaticCatalogModelMock.mockReturnValueOnce(undefined);
+    const cfg = {
+      models: {
+        providers: {
+          xiaomi: {
+            baseUrl: "https://api.xiaomimimo.com/v1",
+            models: [
+              {
+                id: "mimo-v2.5-pro",
+                name: "mimo-v2.5-pro",
+              },
+            ],
+          },
+        },
+      },
+    } as unknown as OpenClawConfig;
+
+    const result = resolveModelForTest("xiaomi", "mimo-v2.5-pro", "/tmp/agent", cfg);
+    const model = expectResolvedModel(result);
+
+    expect(model.id).toBe("mimo-v2.5-pro");
+    expect(model.baseUrl).toBe("https://api.xiaomimimo.com/v1");
+    expect(model.maxTokens).toBeUndefined();
+  });
+
   it("inherits bundled static transport for configured provider fallback models", () => {
     resolveBundledStaticCatalogModelMock.mockReturnValueOnce({
       provider: "deepseek",

src/agents/embedded-agent-runner/model.ts

@@ -1330,6 +1330,11 @@ function resolveConfiguredFallbackModel(params: {
     compat: fallbackCompat,
     reasoning: metadataModel?.reasoning,
   });
+  const resolvedFallbackMaxTokens =
+    configuredModel?.maxTokens ??
+    providerConfig?.maxTokens ??
+    providerConfig?.models?.[0]?.maxTokens ??
+    staticCatalogModel?.maxTokens;
   return normalizeResolvedModel({
     provider,
     cfg,
@@ -1365,12 +1370,11 @@ function resolveConfiguredFallbackModel(params: {
             providerConfig?.contextTokens ??
             providerConfig?.models?.[0]?.contextTokens ??
             staticCatalogModel?.contextTokens,
-          maxTokens:
-            configuredModel?.maxTokens ??
-            providerConfig?.maxTokens ??
-            providerConfig?.models?.[0]?.maxTokens ??
-            staticCatalogModel?.maxTokens ??
-            DEFAULT_CONTEXT_TOKENS,
+          // maxTokens is a wire-level output cap, not a context-budget fallback.
+          // Omit an unknown cap so strict providers can apply their own limit.
+          ...(resolvedFallbackMaxTokens !== undefined
+            ? { maxTokens: resolvedFallbackMaxTokens }
+            : {}),
           ...(resolvedParams ? { params: resolvedParams } : {}),
           ...(requestTimeoutMs !== undefined ? { requestTimeoutMs } : {}),
           headers: requestConfig.headers,

src/agents/openai-transport-stream.test.ts

@@ -7667,6 +7667,31 @@ describe("openai transport stream", () => {
     expect(params).not.toHaveProperty("max_tokens");
   });
 
+  it("omits output-token fields when the resolved model has no known cap", () => {
+    const params = buildOpenAICompletionsParams(
+      {
+        id: "mimo-v2.5-pro",
+        name: "MiMo V2.5 Pro",
+        api: "openai-completions",
+        provider: "xiaomi",
+        baseUrl: "https://api.xiaomimimo.com/v1",
+        reasoning: true,
+        input: ["text"],
+        cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+        contextWindow: 1_048_576,
+      } as Model<"openai-completions">,
+      {
+        systemPrompt: "system",
+        messages: [],
+        tools: [],
+      } as never,
+      undefined,
+    );
+
+    expect(params).not.toHaveProperty("max_completion_tokens");
+    expect(params).not.toHaveProperty("max_tokens");
+  });
+
   it("uses model params max_completion_tokens for OpenAI completions before model maxTokens", () => {
     const params = buildOpenAICompletionsParams(
       {