From 17c9805259f3605dea80022aa7a6f71424548abd Mon Sep 17 00:00:00 2001
From: Naufal Malik Rabbani <mrnaufal17@gmail.com>
Date: Tue, 18 Aug 2026 14:19:34 +0700
Subject: [PATCH] fix(commandcode): strip thinking suffix from params.model

Upstream reads params.model and rejects unknown ids. The client thinking
suffix "model(level)" was copied verbatim into params.model (chatCore
strips only the top-level model field), so requests like
gpt-5.6-luna(max) failed with 403 "Model/provider not recognized:
anthropic:gpt-5.6-luna(max)".

Strip the suffix in the commandcode translator; applyThinking still
consumes the suffix for the thinking override before the body is sent.
---
 .../request/openai-to-commandcode.js          |  6 ++++-
 tests/unit/openai-to-commandcode.test.js      | 26 +++++++++++++++++++
 2 files changed, 31 insertions(+), 1 deletion(-)

diff --git a/open-sse/translator/request/openai-to-commandcode.js b/open-sse/translator/request/openai-to-commandcode.js
index 9825048bd4..b0f79d0bb2 100644
--- a/open-sse/translator/request/openai-to-commandcode.js
+++ b/open-sse/translator/request/openai-to-commandcode.js
@@ -13,6 +13,7 @@ import { register } from "../index.js";
 import { FORMATS } from "../formats.js";
 import { randomUUID } from "crypto";
 import { ROLE, OPENAI_BLOCK } from "../schema/index.js";
+import { stripThinkingSuffix } from "../concerns/thinkingUnified.js";
 import { DEFAULT_MAX_TOKENS } from "../../config/runtimeConfig.js";
 
 function flattenText(content) {
@@ -136,7 +137,10 @@ function convertTools(tools) {
 export function openaiToCommandCodeRequest(model, body, stream /* , credentials */) {
   const { messages, system } = convertMessages(body.messages);
   const params = {
-    model,
+    // Upstream reads params.model and rejects unknown ids — never leak the
+    // client thinking suffix "model(level)" (chatCore strips only the top-level
+    // model field; the suffix is consumed by applyThinking, not the wire).
+    model: stripThinkingSuffix(model),
     messages,
     stream: stream !== false,
     max_tokens: body.max_tokens ?? body.max_output_tokens ?? DEFAULT_MAX_TOKENS,
diff --git a/tests/unit/openai-to-commandcode.test.js b/tests/unit/openai-to-commandcode.test.js
index 7f12dc85f1..27732c4427 100644
--- a/tests/unit/openai-to-commandcode.test.js
+++ b/tests/unit/openai-to-commandcode.test.js
@@ -28,6 +28,32 @@ describe("openaiToCommandCodeRequest — basic envelope", () => {
   });
 });
 
+describe("openaiToCommandCodeRequest — thinking suffix stripping", () => {
+  it("strips a client thinking suffix from params.model (upstream rejects it)", () => {
+    const out = openaiToCommandCodeRequest("gpt-5.6-luna(max)", {
+      messages: [{ role: "user", content: "hi" }],
+    }, true);
+
+    expect(out.params.model).toBe("gpt-5.6-luna");
+  });
+
+  it("strips the suffix from family-prefixed ids too", () => {
+    const out = openaiToCommandCodeRequest("deepseek/deepseek-v4-pro(max)", {
+      messages: [{ role: "user", content: "hi" }],
+    }, true);
+
+    expect(out.params.model).toBe("deepseek/deepseek-v4-pro");
+  });
+
+  it("leaves plain model ids untouched", () => {
+    const out = openaiToCommandCodeRequest("deepseek/deepseek-v4-pro", {
+      messages: [{ role: "user", content: "hi" }],
+    }, true);
+
+    expect(out.params.model).toBe("deepseek/deepseek-v4-pro");
+  });
+});
+
 describe("openaiToCommandCodeRequest — system handling", () => {
   it("hoists system messages to params.system (string), not messages[]", () => {
     const out = openaiToCommandCodeRequest(MODEL, {