fix: map thinkingLevel to reasoning.effort for openai-responses-defaults family

CHANGELOG.md

@@ -33,6 +33,7 @@ Docs: https://docs.openclaw.ai
 - Sessions/reset: clear auto-sourced model, provider, and auth-profile overrides on `/new` and `/reset` while preserving explicit user selections, so channel sessions stop staying pinned to runtime fallback choices. (#69419) Thanks @sk7n4k3d.
 - Sessions/costs: snapshot `estimatedCostUsd` like token counters so repeated persist paths no longer compound the same run cost by up to dozens of times. (#69403) Thanks @MrMiaigi.
 - OpenAI Codex: route ChatGPT/Codex OAuth Responses requests through the `/backend-api/codex` endpoint so `openai-codex/gpt-5.4` no longer hits the removed `/backend-api/responses` alias. (#69336) Thanks @mzogithub.
+- OpenAI/Responses: omit disabled reasoning payloads when `/think off` is active, so GPT reasoning models no longer receive unsupported `reasoning.effort: "none"` requests. (#61982) Thanks @a-tokyo.
 - Gateway/pairing: treat loopback shared-secret node-host, TUI, and gateway clients as local for pairing decisions, so trusted local tools no longer reconnect as remote clients and fail with `pairing required`. (#69431) Thanks @SARAMALI15792.
 - Active Memory: degrade gracefully when memory recall fails during prompt building, logging a warning and letting the reply continue without memory context instead of failing the whole turn. (#69485) Thanks @Magicray1217.
 - Ollama: add provider-policy defaults for `baseUrl` and `models` so implicit local discovery can run before config validation rejects a minimal Ollama provider config. (#69370) Thanks @PratikRai0101.

scripts/check-gateway-watch-regression.mjs

@@ -6,6 +6,7 @@ import net from "node:net";
 import os from "node:os";
 import path from "node:path";
 import process from "node:process";
+import { pathToFileURL } from "node:url";
 import { writeBuildStamp } from "./build-stamp.mjs";
 import { resolveBuildRequirement } from "./run-node.mjs";
 
@@ -180,6 +181,23 @@ function snapshotTree(rootName) {
   return stats;
 }
 
+export function isIgnoredDistRuntimeWatchPath(entry) {
+  return (
+    entry === "dist-runtime/extensions/node_modules" ||
+    entry.startsWith("dist-runtime/extensions/node_modules/")
+  );
+}
+
+function summarizeDistRuntimeAddedPaths(added) {
+  const addedPaths = added.filter((entry) => entry.startsWith("dist-runtime/"));
+  const ignoredDependencyAddedPaths = addedPaths.filter(isIgnoredDistRuntimeWatchPath);
+  const topologyAddedPaths = addedPaths.filter((entry) => !isIgnoredDistRuntimeWatchPath(entry));
+  return {
+    ignoredDependencyAddedPaths,
+    topologyAddedPaths,
+  };
+}
+
 function writeSnapshot(snapshotDir) {
   ensureDir(snapshotDir);
   const pathEntries = [...listTreeEntries("dist"), ...listTreeEntries("dist-runtime")];
@@ -606,11 +624,15 @@ async function main() {
   const post = writeSnapshot(postDir);
   const diff = writeDiffArtifacts(options.outputDir, preDir, postDir);
 
-  const distRuntimeFileGrowth = post.distRuntime.files - pre.distRuntime.files;
-  const distRuntimeByteGrowth = post.distRuntime.apparentBytes - pre.distRuntime.apparentBytes;
-  const distRuntimeAddedPaths = diff.added.filter((entry) =>
-    entry.startsWith("dist-runtime/"),
-  ).length;
+  const distRuntimeAddedPathSummary = summarizeDistRuntimeAddedPaths(diff.added);
+  const distRuntimeAddedPaths = distRuntimeAddedPathSummary.topologyAddedPaths.length;
+  const distRuntimeIgnoredDependencyAddedPaths =
+    distRuntimeAddedPathSummary.ignoredDependencyAddedPaths.length;
+  const distRuntimeFileGrowth = distRuntimeAddedPaths;
+  const distRuntimeByteGrowth =
+    distRuntimeAddedPaths === 0
+      ? 0
+      : post.distRuntime.apparentBytes - pre.distRuntime.apparentBytes;
   const totalCpuMs = Math.round(
     (watchResult.timing.userSeconds + watchResult.timing.sysSeconds) * 1000,
   );
@@ -639,6 +661,7 @@ async function main() {
     distRuntimeByteGrowth,
     distRuntimeByteGrowthMax: options.distRuntimeByteGrowthMax,
     distRuntimeAddedPaths,
+    distRuntimeIgnoredDependencyAddedPaths,
     addedPaths: diff.added.length,
     removedPaths: diff.removed.length,
     watchExit: watchResult.exit,
@@ -699,4 +722,6 @@ async function main() {
   process.exit(0);
 }
 
-await main();
+if (import.meta.url === pathToFileURL(process.argv[1] ?? "").href) {
+  await main();
+}

src/agents/pi-embedded-runner-extraparams.test.ts

@@ -297,6 +297,7 @@ import {
   createOpenAIServiceTierWrapper,
   createOpenAIStringContentWrapper,
   createOpenAITextVerbosityWrapper,
+  createOpenAIThinkingLevelWrapper,
   resolveOpenAIFastMode,
   resolveOpenAIServiceTier,
   resolveOpenAITextVerbosity,
@@ -434,7 +435,9 @@ function createTestOpenAIProviderWrapper(
   });
   streamFn = createOpenAIStringContentWrapper(streamFn);
   return createOpenAIResponsesContextManagementWrapper(
-    createOpenAIReasoningCompatibilityWrapper(streamFn),
+    createOpenAIReasoningCompatibilityWrapper(
+      createOpenAIThinkingLevelWrapper(streamFn, params.context.thinkingLevel),
+    ),
     params.context.extraParams,
   );
 }

src/agents/pi-embedded-runner/openai-stream-wrappers.test.ts

@@ -0,0 +1,195 @@
+import type { StreamFn } from "@mariozechner/pi-agent-core";
+import type { Model } from "@mariozechner/pi-ai";
+import { createAssistantMessageEventStream } from "@mariozechner/pi-ai";
+import { describe, expect, it } from "vitest";
+import { createOpenAIThinkingLevelWrapper } from "./openai-stream-wrappers.js";
+
+function createPayloadCapture(opts?: { initialReasoning?: unknown }) {
+  const payloads: Array<Record<string, unknown>> = [];
+  const baseStreamFn: StreamFn = (model, _context, options) => {
+    const payload: Record<string, unknown> = { model: model.id };
+    if (opts?.initialReasoning !== undefined) {
+      payload.reasoning = structuredClone(opts.initialReasoning);
+    }
+    options?.onPayload?.(payload, model);
+    payloads.push(structuredClone(payload));
+    return createAssistantMessageEventStream();
+  };
+  return { baseStreamFn, payloads };
+}
+
+const codexModel = {
+  api: "openai-codex-responses",
+  provider: "openai-codex",
+  id: "gpt-5.1-codex",
+} as Model<"openai-codex-responses">;
+
+const openaiModel = {
+  api: "openai-responses",
+  provider: "openai",
+  id: "gpt-5.2",
+} as Model<"openai-responses">;
+
+describe("createOpenAIThinkingLevelWrapper", () => {
+  it("overrides effort on reasoning-capable model when thinkingLevel is medium", () => {
+    const { baseStreamFn, payloads } = createPayloadCapture({
+      initialReasoning: { effort: "none" },
+    });
+    const wrapped = createOpenAIThinkingLevelWrapper(baseStreamFn, "medium");
+    void wrapped(codexModel, { messages: [] }, {});
+
+    expect(payloads[0]?.reasoning).toEqual({ effort: "medium" });
+  });
+
+  it("overrides effort on reasoning-capable model when thinkingLevel is high", () => {
+    const { baseStreamFn, payloads } = createPayloadCapture({
+      initialReasoning: { effort: "none" },
+    });
+    const wrapped = createOpenAIThinkingLevelWrapper(baseStreamFn, "high");
+    void wrapped(openaiModel, { messages: [] }, {});
+
+    expect(payloads[0]?.reasoning).toEqual({ effort: "high" });
+  });
+
+  it("removes reasoning when thinkingLevel is off on reasoning-capable model", () => {
+    const { baseStreamFn, payloads } = createPayloadCapture({
+      initialReasoning: { effort: "medium" },
+    });
+    const wrapped = createOpenAIThinkingLevelWrapper(baseStreamFn, "off");
+    void wrapped(codexModel, { messages: [] }, {});
+
+    expect(payloads[0]).not.toHaveProperty("reasoning");
+  });
+
+  it("maps adaptive thinkingLevel to medium effort on reasoning-capable model", () => {
+    const { baseStreamFn, payloads } = createPayloadCapture({
+      initialReasoning: { effort: "none" },
+    });
+    const wrapped = createOpenAIThinkingLevelWrapper(baseStreamFn, "adaptive");
+    void wrapped(codexModel, { messages: [] }, {});
+
+    expect(payloads[0]?.reasoning).toEqual({ effort: "medium" });
+  });
+
+  it("replaces string disabled reasoning when thinkingLevel is enabled", () => {
+    const { baseStreamFn, payloads } = createPayloadCapture({ initialReasoning: "none" });
+    const wrapped = createOpenAIThinkingLevelWrapper(baseStreamFn, "low");
+    void wrapped(codexModel, { messages: [] }, {});
+
+    expect(payloads[0]?.reasoning).toEqual({ effort: "low" });
+  });
+
+  it("does not add reasoning for non-reasoning models without existing reasoning payload", () => {
+    const { baseStreamFn, payloads } = createPayloadCapture();
+    const wrapped = createOpenAIThinkingLevelWrapper(baseStreamFn, "medium");
+    void wrapped(openaiModel, { messages: [] }, {});
+
+    expect(payloads[0]?.reasoning).toBeUndefined();
+  });
+
+  it("overrides existing reasoning.effort from upstream wrappers", () => {
+    const baseStreamFn: StreamFn = (model, _context, options) => {
+      const payload: Record<string, unknown> = {
+        model: model.id,
+        reasoning: { effort: "none" },
+      };
+      options?.onPayload?.(payload, model);
+      return createAssistantMessageEventStream();
+    };
+
+    const payloads: Array<Record<string, unknown>> = [];
+    const capture: StreamFn = (model, context, options) => {
+      return baseStreamFn(model, context, {
+        ...options,
+        onPayload: (payload, m) => {
+          options?.onPayload?.(payload, m);
+          payloads.push(structuredClone(payload as Record<string, unknown>));
+        },
+      });
+    };
+
+    const wrapped = createOpenAIThinkingLevelWrapper(capture, "medium");
+    void wrapped(codexModel, { messages: [] }, {});
+
+    expect(payloads[0]?.reasoning).toEqual({ effort: "medium" });
+  });
+
+  it("returns underlying streamFn unchanged when thinkingLevel is undefined", () => {
+    const { baseStreamFn } = createPayloadCapture();
+    const wrapped = createOpenAIThinkingLevelWrapper(baseStreamFn, undefined);
+    expect(wrapped).toBe(baseStreamFn);
+  });
+
+  it("preserves other reasoning properties when overriding effort", () => {
+    const baseStreamFn: StreamFn = (model, _context, options) => {
+      const payload: Record<string, unknown> = {
+        model: model.id,
+        reasoning: { effort: "none", summary: "auto" },
+      };
+      options?.onPayload?.(payload, model);
+      return createAssistantMessageEventStream();
+    };
+
+    const payloads: Array<Record<string, unknown>> = [];
+    const capture: StreamFn = (model, context, options) => {
+      return baseStreamFn(model, context, {
+        ...options,
+        onPayload: (payload, m) => {
+          options?.onPayload?.(payload, m);
+          payloads.push(structuredClone(payload as Record<string, unknown>));
+        },
+      });
+    };
+
+    const wrapped = createOpenAIThinkingLevelWrapper(capture, "high");
+    void wrapped(codexModel, { messages: [] }, {});
+
+    expect(payloads[0]?.reasoning).toEqual({ effort: "high", summary: "auto" });
+  });
+
+  it("does not inject reasoning for completions API on proxy routes", () => {
+    const { baseStreamFn, payloads } = createPayloadCapture();
+    const wrapped = createOpenAIThinkingLevelWrapper(baseStreamFn, "medium");
+    void wrapped(
+      {
+        api: "openai-completions",
+        provider: "openai",
+        id: "gpt-4o",
+        baseUrl: "https://proxy.example.com/v1",
+      } as Model<"openai-completions">,
+      { messages: [] },
+      {},
+    );
+
+    expect(payloads[0]?.reasoning).toBeUndefined();
+  });
+
+  it("does not inject reasoning for proxy routes with custom baseUrl", () => {
+    const { baseStreamFn, payloads } = createPayloadCapture();
+    const wrapped = createOpenAIThinkingLevelWrapper(baseStreamFn, "medium");
+    void wrapped(
+      {
+        api: "openai-responses",
+        provider: "openai",
+        id: "gpt-5.2",
+        baseUrl: "https://proxy.example.com/v1",
+      } as Model<"openai-responses">,
+      { messages: [] },
+      {},
+    );
+
+    expect(payloads[0]?.reasoning).toBeUndefined();
+  });
+
+  it("passes through all thinking levels correctly on reasoning-capable models", () => {
+    const levels = ["minimal", "low", "medium", "high", "xhigh"] as const;
+    for (const level of levels) {
+      const { baseStreamFn, payloads } = createPayloadCapture({
+        initialReasoning: { effort: "none" },
+      });
+      const wrapped = createOpenAIThinkingLevelWrapper(baseStreamFn, level);
+      void wrapped(codexModel, { messages: [] }, {});
+      expect(payloads[0]?.reasoning).toEqual({ effort: level });
+    }
+  });
+});

src/agents/pi-embedded-runner/openai-stream-wrappers.ts

@@ -1,6 +1,7 @@
 import type { StreamFn } from "@mariozechner/pi-agent-core";
 import type { SimpleStreamOptions } from "@mariozechner/pi-ai";
 import { streamSimple } from "@mariozechner/pi-ai";
+import type { ThinkLevel } from "../../auto-reply/thinking.js";
 import type { OpenClawConfig } from "../../config/types.openclaw.js";
 import { normalizeOptionalLowercaseString, readStringValue } from "../../shared/string-coerce.js";
 import {
@@ -15,6 +16,7 @@ import {
 import { resolveOpenAITextVerbosity, type OpenAITextVerbosity } from "../openai-text-verbosity.js";
 import { resolveProviderRequestPolicyConfig } from "../provider-request-config.js";
 import { log } from "./logger.js";
+import { mapThinkingLevelToReasoningEffort } from "./reasoning-effort-utils.js";
 import { streamWithPayloadPatch } from "./stream-payload-utils.js";
 
 type OpenAIServiceTier = "auto" | "default" | "flex" | "priority";
@@ -224,6 +226,43 @@ export function createOpenAIStringContentWrapper(baseStreamFn: StreamFn | undefi
   };
 }
 
+export function createOpenAIThinkingLevelWrapper(
+  baseStreamFn: StreamFn | undefined,
+  thinkingLevel?: ThinkLevel,
+): StreamFn {
+  const underlying = baseStreamFn ?? streamSimple;
+  if (!thinkingLevel) {
+    return underlying;
+  }
+  return (model, context, options) => {
+    if (!shouldApplyOpenAIReasoningCompatibility(model)) {
+      return underlying(model, context, options);
+    }
+    return streamWithPayloadPatch(underlying, model, context, options, (payloadObj) => {
+      const existingReasoning = payloadObj.reasoning;
+      if (thinkingLevel === "off") {
+        if (existingReasoning !== undefined) {
+          delete payloadObj.reasoning;
+        }
+        return;
+      }
+
+      if (existingReasoning === "none") {
+        payloadObj.reasoning = { effort: mapThinkingLevelToReasoningEffort(thinkingLevel) };
+        return;
+      }
+      if (
+        existingReasoning &&
+        typeof existingReasoning === "object" &&
+        !Array.isArray(existingReasoning)
+      ) {
+        (existingReasoning as Record<string, unknown>).effort =
+          mapThinkingLevelToReasoningEffort(thinkingLevel);
+      }
+    });
+  };
+}
+
 export function createOpenAIFastModeWrapper(baseStreamFn: StreamFn | undefined): StreamFn {
   const underlying = baseStreamFn ?? streamSimple;
   return (model, context, options) => {

src/agents/pi-embedded-runner/proxy-stream-wrappers.ts

@@ -7,6 +7,7 @@ import { resolveProviderRequestPolicy } from "../provider-attribution.js";
 import { resolveProviderRequestPolicyConfig } from "../provider-request-config.js";
 import { applyAnthropicEphemeralCacheControlMarkers } from "./anthropic-cache-control-payload.js";
 import { isAnthropicModelRef } from "./anthropic-family-cache-semantics.js";
+import { mapThinkingLevelToReasoningEffort } from "./reasoning-effort-utils.js";
 import { streamWithPayloadPatch } from "./stream-payload-utils.js";
 const KILOCODE_FEATURE_HEADER = "X-KILOCODE-FEATURE";
 const KILOCODE_FEATURE_DEFAULT = "openclaw";
@@ -17,21 +18,6 @@ function resolveKilocodeAppHeaders(): Record<string, string> {
   return { [KILOCODE_FEATURE_HEADER]: feature };
 }
 
-function mapThinkingLevelToOpenRouterReasoningEffort(
-  thinkingLevel: ThinkLevel,
-): "none" | "minimal" | "low" | "medium" | "high" | "xhigh" {
-  if (thinkingLevel === "off") {
-    return "none";
-  }
-  if (thinkingLevel === "adaptive") {
-    return "medium";
-  }
-  if (thinkingLevel === "max") {
-    return "xhigh";
-  }
-  return thinkingLevel;
-}
-
 function normalizeProxyReasoningPayload(payload: unknown, thinkingLevel?: ThinkLevel): void {
   if (!payload || typeof payload !== "object") {
     return;
@@ -51,11 +37,11 @@ function normalizeProxyReasoningPayload(payload: unknown, thinkingLevel?: ThinkL
   ) {
     const reasoningObj = existingReasoning as Record<string, unknown>;
     if (!("max_tokens" in reasoningObj) && !("effort" in reasoningObj)) {
-      reasoningObj.effort = mapThinkingLevelToOpenRouterReasoningEffort(thinkingLevel);
+      reasoningObj.effort = mapThinkingLevelToReasoningEffort(thinkingLevel);
     }
   } else if (!existingReasoning) {
     payloadObj.reasoning = {
-      effort: mapThinkingLevelToOpenRouterReasoningEffort(thinkingLevel),
+      effort: mapThinkingLevelToReasoningEffort(thinkingLevel),
     };
   }
 }

src/agents/pi-embedded-runner/reasoning-effort-utils.test.ts

@@ -0,0 +1,23 @@
+import { describe, expect, it } from "vitest";
+import { mapThinkingLevelToReasoningEffort } from "./reasoning-effort-utils.js";
+
+describe("mapThinkingLevelToReasoningEffort", () => {
+  it('maps "off" to "none"', () => {
+    expect(mapThinkingLevelToReasoningEffort("off")).toBe("none");
+  });
+
+  it('maps "adaptive" to "medium"', () => {
+    expect(mapThinkingLevelToReasoningEffort("adaptive")).toBe("medium");
+  });
+
+  it('maps "max" to "xhigh"', () => {
+    expect(mapThinkingLevelToReasoningEffort("max")).toBe("xhigh");
+  });
+
+  it.each(["minimal", "low", "medium", "high", "xhigh"] as const)(
+    "passes through %s unchanged",
+    (level) => {
+      expect(mapThinkingLevelToReasoningEffort(level)).toBe(level);
+    },
+  );
+});

src/agents/pi-embedded-runner/reasoning-effort-utils.ts

@@ -0,0 +1,16 @@
+import type { ThinkLevel } from "../../auto-reply/thinking.js";
+
+export type ReasoningEffort = "none" | "minimal" | "low" | "medium" | "high" | "xhigh";
+
+export function mapThinkingLevelToReasoningEffort(thinkingLevel: ThinkLevel): ReasoningEffort {
+  if (thinkingLevel === "off") {
+    return "none";
+  }
+  if (thinkingLevel === "adaptive") {
+    return "medium";
+  }
+  if (thinkingLevel === "max") {
+    return "xhigh";
+  }
+  return thinkingLevel;
+}

src/plugin-sdk/provider-stream.ts

@@ -11,7 +11,9 @@ import {
   createOpenAIReasoningCompatibilityWrapper,
   createOpenAIResponsesContextManagementWrapper,
   createOpenAIServiceTierWrapper,
+  createOpenAIStringContentWrapper,
   createOpenAITextVerbosityWrapper,
+  createOpenAIThinkingLevelWrapper,
   resolveOpenAIFastMode,
   resolveOpenAIServiceTier,
   resolveOpenAITextVerbosity,
@@ -118,8 +120,11 @@ export function buildProviderStreamFamilyHooks(
             config: ctx.config,
             agentDir: ctx.agentDir,
           });
+          nextStreamFn = createOpenAIStringContentWrapper(nextStreamFn);
           return createOpenAIResponsesContextManagementWrapper(
-            createOpenAIReasoningCompatibilityWrapper(nextStreamFn),
+            createOpenAIReasoningCompatibilityWrapper(
+              createOpenAIThinkingLevelWrapper(nextStreamFn, ctx.thinkingLevel),
+            ),
             ctx.extraParams,
           );
         },

test/scripts/check-gateway-watch-regression.test.ts

@@ -0,0 +1,22 @@
+import { describe, expect, it } from "vitest";
+import { isIgnoredDistRuntimeWatchPath } from "../../scripts/check-gateway-watch-regression.mjs";
+
+describe("check-gateway-watch-regression", () => {
+  it("ignores top-level dist-runtime extension dependency repairs", () => {
+    expect(isIgnoredDistRuntimeWatchPath("dist-runtime/extensions/node_modules")).toBe(true);
+    expect(
+      isIgnoredDistRuntimeWatchPath(
+        "dist-runtime/extensions/node_modules/playwright-core/index.js",
+      ),
+    ).toBe(true);
+  });
+
+  it("keeps plugin runtime graph paths counted", () => {
+    expect(isIgnoredDistRuntimeWatchPath("dist-runtime/extensions/openai/index.js")).toBe(false);
+    expect(
+      isIgnoredDistRuntimeWatchPath(
+        "dist-runtime/extensions/openai/node_modules/openclaw/index.js",
+      ),
+    ).toBe(false);
+  });
+});