From 99cfb3dab26a278dd3ea89a413c4a37f971ece9d Mon Sep 17 00:00:00 2001
From: Robby <robbyczgw@gmail.com>
Date: Sun, 22 Feb 2026 18:14:12 +0100
Subject: [PATCH] fix(openrouter): pass reasoning.effort based on thinking
 level (#14664) (#17236)

* fix(openrouter): pass reasoning.effort to OpenRouter API (#14664)

* Agents: pass thinkLevel to extra-params wrapper

* Changelog: note fix/openrouter-reasoning-effort-14664 OpenRouter fix

* Changelog: fix OpenRouter entry text

---------

Co-authored-by: Vincent Koc <vincentkoc@ieee.org>
---
 CHANGELOG.md                                  |  1 +
 src/agents/pi-embedded-runner/extra-params.ts | 57 +++++++++++++++++--
 src/agents/pi-embedded-runner/run/attempt.ts  |  1 +
 3 files changed, 53 insertions(+), 6 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index fb3633ea9..d6f5eeb7d 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -26,6 +26,7 @@ Docs: https://docs.openclaw.ai
 
 ### Fixes
 
+- Providers/OpenRouter: map `/think` levels to `reasoning.effort` in embedded runs while preserving explicit `reasoning.max_tokens` payloads. (#17236) Thanks @robbyczgw-cla.
 - Gateway/OpenRouter: preserve stored session provider when model IDs are vendor-prefixed (for example, `anthropic/...`) so follow-up turns do not incorrectly route to direct provider APIs. (#22753) Thanks @dndodson.
 - Providers/OpenRouter: preserve the required `openrouter/` prefix for OpenRouter-native model IDs during model-ref normalization. (#12942) Thanks @omair445.
 - Providers/OpenRouter: pass through provider routing parameters from model params.provider to OpenRouter request payloads for provider selection controls. (#17148) Thanks @carrotRakko.
diff --git a/src/agents/pi-embedded-runner/extra-params.ts b/src/agents/pi-embedded-runner/extra-params.ts
index 9b5587b9d..df0812c67 100644
--- a/src/agents/pi-embedded-runner/extra-params.ts
+++ b/src/agents/pi-embedded-runner/extra-params.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/config.js";
 import { log } from "./logger.js";
 
@@ -290,19 +291,62 @@ function createAnthropicBetaHeadersWrapper(
 }
 
 /**
- * Create a streamFn wrapper that adds OpenRouter app attribution headers.
- * These headers allow OpenClaw to appear on OpenRouter's leaderboard.
+ * Map OpenClaw's ThinkLevel to OpenRouter's reasoning.effort values.
+ * "off" maps to "none"; all other levels pass through as-is.
  */
-function createOpenRouterHeadersWrapper(baseStreamFn: StreamFn | undefined): StreamFn {
+function mapThinkingLevelToOpenRouterReasoningEffort(
+  thinkingLevel: ThinkLevel,
+): "none" | "minimal" | "low" | "medium" | "high" | "xhigh" {
+  if (thinkingLevel === "off") {
+    return "none";
+  }
+  return thinkingLevel;
+}
+
+/**
+ * Create a streamFn wrapper that adds OpenRouter app attribution headers
+ * and injects reasoning.effort based on the configured thinking level.
+ */
+function createOpenRouterWrapper(
+  baseStreamFn: StreamFn | undefined,
+  thinkingLevel?: ThinkLevel,
+): StreamFn {
   const underlying = baseStreamFn ?? streamSimple;
-  return (model, context, options) =>
-    underlying(model, context, {
+  return (model, context, options) => {
+    const onPayload = options?.onPayload;
+    return underlying(model, context, {
       ...options,
       headers: {
         ...OPENROUTER_APP_HEADERS,
         ...options?.headers,
       },
+      onPayload: (payload) => {
+        if (thinkingLevel && payload && typeof payload === "object") {
+          const payloadObj = payload as Record<string, unknown>;
+          const existingReasoning = payloadObj.reasoning;
+
+          // OpenRouter treats reasoning.effort and reasoning.max_tokens as
+          // alternative controls. If max_tokens is already present, do not
+          // inject effort and do not overwrite caller-supplied reasoning.
+          if (
+            existingReasoning &&
+            typeof existingReasoning === "object" &&
+            !Array.isArray(existingReasoning)
+          ) {
+            const reasoningObj = existingReasoning as Record<string, unknown>;
+            if (!("max_tokens" in reasoningObj) && !("effort" in reasoningObj)) {
+              reasoningObj.effort = mapThinkingLevelToOpenRouterReasoningEffort(thinkingLevel);
+            }
+          } else if (!existingReasoning) {
+            payloadObj.reasoning = {
+              effort: mapThinkingLevelToOpenRouterReasoningEffort(thinkingLevel),
+            };
+          }
+        }
+        onPayload?.(payload);
+      },
     });
+  };
 }
 
 /**
@@ -350,6 +394,7 @@ export function applyExtraParamsToAgent(
   provider: string,
   modelId: string,
   extraParamsOverride?: Record<string, unknown>,
+  thinkingLevel?: ThinkLevel,
 ): void {
   const extraParams = resolveExtraParams({
     cfg,
@@ -380,7 +425,7 @@ export function applyExtraParamsToAgent(
 
   if (provider === "openrouter") {
     log.debug(`applying OpenRouter app attribution headers for ${provider}/${modelId}`);
-    agent.streamFn = createOpenRouterHeadersWrapper(agent.streamFn);
+    agent.streamFn = createOpenRouterWrapper(agent.streamFn, thinkingLevel);
   }
 
   // Enable Z.AI tool_stream for real-time tool call streaming.
diff --git a/src/agents/pi-embedded-runner/run/attempt.ts b/src/agents/pi-embedded-runner/run/attempt.ts
index 383d810e7..d8ba48e15 100644
--- a/src/agents/pi-embedded-runner/run/attempt.ts
+++ b/src/agents/pi-embedded-runner/run/attempt.ts
@@ -736,6 +736,7 @@ export async function runEmbeddedAttempt(
         params.provider,
         params.modelId,
         params.streamParams,
+        params.thinkLevel,
       );
 
       if (cacheTrace) {