From 288940960a151bfaff027903942c13764043ea09 Mon Sep 17 00:00:00 2001
From: thienpv <thien.pv@shb.com.vn>
Date: Thu, 9 Jul 2026 15:10:53 +0700
Subject: [PATCH] fix(translator): clamp thinking effort max->xhigh for OpenAI
 format (#2466)

Claude Code sends reasoning_effort "max" (its top level); OpenAI enum caps
at "xhigh" and rejects "max" with HTTP 400 "max effort not support".
applyFormat case "openai" now clamps "max"->"xhigh" before assigning
body.reasoning_effort; other levels pass through unchanged.

Add regression test covering client output_config.effort, direct
reasoning_effort, passthrough of xhigh/high, and budget_tokens capping.

Co-authored-by: Cursor <cursoragent@cursor.com>
---
 .../translator/concerns/thinkingUnified.js    |  3 +-
 .../thinking-effort-openai-max-clamp.test.js  | 40 +++++++++++++++++++
 2 files changed, 42 insertions(+), 1 deletion(-)
 create mode 100644 tests/unit/thinking-effort-openai-max-clamp.test.js

diff --git a/open-sse/translator/concerns/thinkingUnified.js b/open-sse/translator/concerns/thinkingUnified.js
index 9e9235de..b8540468 100644
--- a/open-sse/translator/concerns/thinkingUnified.js
+++ b/open-sse/translator/concerns/thinkingUnified.js
@@ -184,7 +184,8 @@ function applyFormat(fmt, body, cfg, caps) {
     case "openai": {
       if (none && canDisable) { body.reasoning_effort = "none"; break; }
       const level = toLevel(eff);
-      if (level) body.reasoning_effort = level;
+      // OpenAI reasoning_effort enum caps at "xhigh" (no "max"); clamp Claude Code's "max".
+      if (level) body.reasoning_effort = level === "max" ? "xhigh" : level;
       break;
     }
     case "claude-adaptive": {
diff --git a/tests/unit/thinking-effort-openai-max-clamp.test.js b/tests/unit/thinking-effort-openai-max-clamp.test.js
new file mode 100644
index 00000000..23917fd1
--- /dev/null
+++ b/tests/unit/thinking-effort-openai-max-clamp.test.js
@@ -0,0 +1,40 @@
+import { describe, expect, it } from "vitest";
+import { applyThinking } from "../../open-sse/translator/concerns/thinkingUnified.js";
+import { FORMATS } from "../../open-sse/translator/formats.js";
+
+// Regression: Claude Code sends thinking effort "max" (its top level). When
+// 9router routes to an OpenAI-format provider, applyThinking() case "openai"
+// must clamp "max"→"xhigh" because OpenAI's reasoning_effort enum has no "max"
+// (L.openai caps at "xhigh"). Without the clamp, upstream returns HTTP 400
+// "max effort not support". See open-sse/providers/thinkingLevels.js:10.
+describe("applyThinking (openai): clamp max effort to xhigh", () => {
+  it("client output_config.effort:\"max\" → reasoning_effort:\"xhigh\" (not \"max\")", () => {
+    const body = { output_config: { effort: "max" } };
+    const out = applyThinking(FORMATS.OPENAI, "gpt-5", body, "openai");
+    expect(out.reasoning_effort).toBe("xhigh");
+  });
+
+  it("direct reasoning_effort:\"max\" clamped to \"xhigh\"", () => {
+    const body = { reasoning_effort: "max" };
+    const out = applyThinking(FORMATS.OPENAI, "gpt-5", body, "openai");
+    expect(out.reasoning_effort).toBe("xhigh");
+  });
+
+  it("\"xhigh\" passes through unchanged (highest valid OpenAI level)", () => {
+    const body = { reasoning_effort: "xhigh" };
+    const out = applyThinking(FORMATS.OPENAI, "gpt-5", body, "openai");
+    expect(out.reasoning_effort).toBe("xhigh");
+  });
+
+  it("\"high\" passes through unchanged", () => {
+    const body = { reasoning_effort: "high" };
+    const out = applyThinking(FORMATS.OPENAI, "gpt-5", body, "openai");
+    expect(out.reasoning_effort).toBe("high");
+  });
+
+  it("max budget (thinking.budget_tokens:128000) → reasoning_effort:\"xhigh\" (budgetToLevel caps at xhigh)", () => {
+    const body = { thinking: { type: "enabled", budget_tokens: 128000 } };
+    const out = applyThinking(FORMATS.OPENAI, "gpt-5", body, "openai");
+    expect(out.reasoning_effort).toBe("xhigh");
+  });
+});
\ No newline at end of file