From 4431f2fec5f12bdeb4d71f6de462b8032f07f6e6 Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Fri, 26 Jun 2026 09:20:57 +0000
Subject: [PATCH 01/19] =?UTF-8?q?fix(Dockerfile):=20=E7=A7=BB=E9=99=A4?=
 =?UTF-8?q?=E4=B8=8D=E5=AE=89=E5=85=A8=E5=A5=97=E4=BB=B6=E6=86=91=E8=AD=89?=
 =?UTF-8?q?=E5=BF=BD=E7=95=A5=E4=B8=A6=E9=87=98=E9=81=B8=20Alpine?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 Dockerfile | 11 ++++-------
 1 file changed, 4 insertions(+), 7 deletions(-)

diff --git a/Dockerfile b/Dockerfile
index d708574..911b6b9 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -6,18 +6,15 @@
 # 更新日期：2026/06/26 11:34:46
 # =============================================================================
 
-# 指定基底映像為 Alpine Linux 最新版；Alpine 體積小，可縮小最終映像大小並加快拉取速度。
-# 需人工確認：使用 latest tag 會在不同時間建置出不同基底版本，可能影響可重現性，
-#             建議釘選明確版本（例如 alpine:3.20）以確保建置一致。
-FROM alpine:latest
+# 指定支援中的 Alpine Linux 固定版本；Alpine 體積小，可縮小最終映像大小並加快拉取速度。
+# 釘選明確版本可避免 latest tag 隨時間變動，提升建置可重現性。
+FROM alpine:3.24.1
 
 # 安裝必要的工具
 # 安裝執行 code review 所需的工具：bash（執行 entrypoint 腳本）、git（前置遠端驗證／取得 diff）、
 # nodejs 與 npm（執行 app 內的 Node.js 程式）。
 # --no-cache：不保留 apk 套件索引快取，避免殘留在映像層中以減少映像大小。
-# 需人工確認：--no-check-certificate 會略過套件來源的憑證驗證，存在中間人攻擊風險，
-#             僅在內網或憑證受限環境下使用；正式環境建議移除以維持安全性。
-RUN apk add --no-cache --no-check-certificate bash git nodejs npm
+RUN apk add --no-cache bash git nodejs npm
 
 # 將專案的 app/ 目錄複製到映像內的 /app；包含 Node.js 程式碼與 package.json 等相依宣告。
 COPY ./app /app
-- 
2.53.0


From bc0df064457ead64223a4d4f3570f2cd56b92d96 Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Fri, 26 Jun 2026 09:20:57 +0000
Subject: [PATCH 02/19] =?UTF-8?q?fix(TLS=20=E9=A9=97=E8=AD=89):=20?=
 =?UTF-8?q?=E5=A5=97=E7=94=A8=E5=85=A8=E5=9F=9F=E6=86=91=E8=AD=89=E5=BF=BD?=
 =?UTF-8?q?=E7=95=A5=E8=A8=AD=E5=AE=9A?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 app/config.js    | 12 ++++++++----
 app/git.js       |  8 +++++++-
 app/gitea.js     |  5 ++---
 app/preflight.js |  4 ++--
 app/usage.js     |  2 ++
 5 files changed, 21 insertions(+), 10 deletions(-)

diff --git a/app/config.js b/app/config.js
index 3f205d1..c92f2b1 100644
--- a/app/config.js
+++ b/app/config.js
@@ -1,5 +1,7 @@
 import https from 'https';
 
+process.env.NODE_TLS_REJECT_UNAUTHORIZED = '0';
+
 export const GITEA_TOKEN = process.env.GITEA_TOKEN || '';
 export const GITEA_COMMENT_TOKEN = process.env.GITEA_COMMENT_TOKEN || '';
 export const GITEA_SERVER_URL = process.env.GITEA_SERVER_URL || 'https://gitea.com';
@@ -14,18 +16,20 @@ export const EXCLUSIONS_PATH = '.gitea/ai-review/exclusions.json';
 
 /**
  * 建立一個停用 TLS 憑證驗證（`rejectUnauthorized: false`）的 HTTPS Agent，
- * 供連接使用自簽或無效憑證的 OpenCode 服務時使用。
+ * 供連接使用自簽或無效憑證的內部服務時使用。
  *
  * @remarks 首次呼叫時建立，之後快取為模組層級單例（singleton）重複使用，
  *          避免每次都新建 Agent 與連線池、浪費 TCP 三次握手。
  *          停用憑證驗證有中間人攻擊風險，僅限受信任的內部環境使用。
  * @returns {import('https').Agent} 已關閉憑證驗證的 HTTPS Agent 單例。
  */
-let _openCodeHttpsAgent = null;
-export function getOpenCodeHttpsAgent() {
-  return (_openCodeHttpsAgent ??= new https.Agent({ rejectUnauthorized: false }));
+let _insecureHttpsAgent = null;
+export function getInsecureHttpsAgent() {
+  return (_insecureHttpsAgent ??= new https.Agent({ rejectUnauthorized: false }));
 }
 
+export const getOpenCodeHttpsAgent = getInsecureHttpsAgent;
+
 /**
  * 依環境變數解析並回傳 LLM 提供者設定。
  *
diff --git a/app/git.js b/app/git.js
index 8fc9423..ea8146e 100644
--- a/app/git.js
+++ b/app/git.js
@@ -61,7 +61,13 @@ function makeRunner(spawn) {
 function withAskpass(workspace, fn, token = GITEA_TOKEN) {
   const askpassScript = path.join(workspace, '.git-askpass.sh');
   fs.writeFileSync(askpassScript, '#!/bin/sh\necho "$GIT_TOKEN"\n', { mode: 0o700 });
-  const credEnv = { ...process.env, GIT_ASKPASS: askpassScript, GIT_USERNAME: 'x-token', GIT_TOKEN: token };
+  const credEnv = {
+    ...process.env,
+    GIT_ASKPASS: askpassScript,
+    GIT_SSL_NO_VERIFY: 'true',
+    GIT_USERNAME: 'x-token',
+    GIT_TOKEN: token,
+  };
   const cleanup = () => { try { fs.unlinkSync(askpassScript); } catch {} };
   let result;
   try {
diff --git a/app/gitea.js b/app/gitea.js
index ccd6b49..3f06b0a 100644
--- a/app/gitea.js
+++ b/app/gitea.js
@@ -1,9 +1,8 @@
 import axios from 'axios';
-import https from 'https';
-import { GITEA_TOKEN, GITEA_COMMENT_TOKEN, GITEA_SERVER_URL, GITEA_REPOSITORY, PR_NUMBER, PR_HEAD_SHA, PR_HEAD_BRANCH } from './config.js';
+import { GITEA_TOKEN, GITEA_COMMENT_TOKEN, GITEA_SERVER_URL, GITEA_REPOSITORY, PR_NUMBER, PR_HEAD_SHA, PR_HEAD_BRANCH, getInsecureHttpsAgent } from './config.js';
 import { line, warn } from './log.js';
 
-const httpsAgent = new https.Agent({ rejectUnauthorized: false });
+const httpsAgent = getInsecureHttpsAgent();
 /**
  * 產生呼叫 Gitea API 所需的 HTTP headers（含 Gitea token 授權與 JSON content-type）。
  * 授權格式為 Gitea 專用的 `token <token>`，並非 OAuth Bearer。
diff --git a/app/preflight.js b/app/preflight.js
index 3f94fb6..90b8814 100644
--- a/app/preflight.js
+++ b/app/preflight.js
@@ -1,18 +1,18 @@
 import axios from 'axios';
-import https from 'https';
 import {
   GITEA_TOKEN,
   GITEA_COMMENT_TOKEN,
   GITEA_SERVER_URL,
   GITEA_REPOSITORY,
   PR_NUMBER,
+  getInsecureHttpsAgent,
   getOpenCodeHttpsAgent,
   getLLMConfig,
 } from './config.js';
 import { verifyRemoteAccess } from './git.js';
 import { step, line, ok, error, result } from './log.js';
 
-const httpsAgent = new https.Agent({ rejectUnauthorized: false });
+const httpsAgent = getInsecureHttpsAgent();
 /**
  * 組出 Gitea REST API v1 的完整網址。
  *
diff --git a/app/usage.js b/app/usage.js
index 9376b56..97eb2df 100644
--- a/app/usage.js
+++ b/app/usage.js
@@ -1,4 +1,5 @@
 import axios from 'axios';
+import { getInsecureHttpsAgent } from './config.js';
 import { warn } from './log.js';
 
 /** 本次執行的 token 累計（跨所有 LLM 呼叫）。 */
@@ -176,6 +177,7 @@ async function fetchOpenRouterQuota({ apiKey, baseURL }, get) {
   const resp = await get(`${stripSlash(baseURL)}/auth/key`, {
     headers: { Authorization: `Bearer ${apiKey}` },
     timeout: 30000,
+    httpsAgent: getInsecureHttpsAgent(),
   });
   const d = resp.data?.data || {};
   const used = num(d.usage);
-- 
2.53.0


From 2fe03e94f03fb06925ca0846ee4887f7f0295c88 Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Fri, 26 Jun 2026 09:20:57 +0000
Subject: [PATCH 03/19] =?UTF-8?q?test(TLS=20=E9=A9=97=E8=AD=89):=20?=
 =?UTF-8?q?=E8=A3=9C=E4=B8=8A=E6=86=91=E8=AD=89=E5=BF=BD=E7=95=A5=E8=A1=8C?=
 =?UTF-8?q?=E7=82=BA=E6=B8=AC=E8=A9=A6?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 app/test/config.test.js | 4 ++++
 app/test/git.test.js    | 3 +++
 app/test/usage.test.js  | 1 +
 3 files changed, 8 insertions(+)

diff --git a/app/test/config.test.js b/app/test/config.test.js
index c6aa636..7139dae 100644
--- a/app/test/config.test.js
+++ b/app/test/config.test.js
@@ -48,4 +48,8 @@ describe('getLLMConfig', () => {
     assert.equal(agent.options.rejectUnauthorized, false);
   });
 
+  it('disables Node TLS certificate verification globally', () => {
+    assert.equal(process.env.NODE_TLS_REJECT_UNAUTHORIZED, '0');
+  });
+
 });
diff --git a/app/test/git.test.js b/app/test/git.test.js
index 9d9b151..9676d06 100644
--- a/app/test/git.test.js
+++ b/app/test/git.test.js
@@ -96,6 +96,7 @@ describe('commitAndPush', () => {
 
     for (const { args, opts } of networkCalls) {
       assert.ok(opts?.env?.GIT_ASKPASS, `GIT_ASKPASS missing for git ${args[0]}`);
+      assert.equal(opts.env.GIT_SSL_NO_VERIFY, 'true', `GIT_SSL_NO_VERIFY missing for git ${args[0]}`);
     }
   });
 
@@ -266,6 +267,7 @@ describe('cloneRepo', () => {
     assert.ok(networkCalls.length > 0, 'expected at least one network git call');
     for (const { args, opts } of networkCalls) {
       assert.ok(opts?.env?.GIT_ASKPASS, `GIT_ASKPASS missing for git ${args[0]}`);
+      assert.equal(opts.env.GIT_SSL_NO_VERIFY, 'true', `GIT_SSL_NO_VERIFY missing for git ${args[0]}`);
     }
   });
 
@@ -308,6 +310,7 @@ describe('verifyRemoteAccess', () => {
     const lsRemote = calls.find(c => c.args[0] === 'ls-remote');
     assert.ok(lsRemote, 'expected git ls-remote to run');
     assert.ok(lsRemote.opts?.env?.GIT_ASKPASS, 'expected GIT_ASKPASS env for ls-remote');
+    assert.equal(lsRemote.opts.env.GIT_SSL_NO_VERIFY, 'true');
   });
 
   it('does not leak the token in ls-remote args', () => {
diff --git a/app/test/usage.test.js b/app/test/usage.test.js
index 9c42bf0..d6b5f4e 100644
--- a/app/test/usage.test.js
+++ b/app/test/usage.test.js
@@ -82,6 +82,7 @@ describe('fetchAccountQuota', () => {
     const get = async (url, opts) => {
       assert.match(url, /openrouter\.ai\/api\/v1\/auth\/key$/);
       assert.equal(opts.headers.Authorization, 'Bearer sk-or-xxx');
+      assert.equal(opts.httpsAgent.options.rejectUnauthorized, false);
       return { data: { data: { usage: 12.4, limit: 100, limit_remaining: 87.6 } } };
     };
     const q = await fetchAccountQuota('openai', { apiKeys: ['sk-or-xxx'], baseURL: 'https://openrouter.ai/api/v1' }, { get });
-- 
2.53.0


From a62bfbd4b8f67641f75e58fdfda2c761c59d4bdb Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Fri, 26 Jun 2026 09:20:57 +0000
Subject: [PATCH 04/19] =?UTF-8?q?chore(ai-review=20=E7=8B=80=E6=85=8B):=20?=
 =?UTF-8?q?=E7=A7=BB=E9=99=A4=E5=B7=B2=E8=A7=A3=E6=B1=BA=20findings?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .gitea/ai-review/findings.json | 35 +---------------------------------
 1 file changed, 1 insertion(+), 34 deletions(-)

diff --git a/.gitea/ai-review/findings.json b/.gitea/ai-review/findings.json
index 1320606..fe51488 100644
--- a/.gitea/ai-review/findings.json
+++ b/.gitea/ai-review/findings.json
@@ -1,34 +1 @@
-[
-  {
-    "level": "critical",
-    "role": "Assassin",
-    "location": "Dockerfile:17",
-    "problem": "在 `apk add` 中使用了 `--no-check-certificate` 參數。這會強制忽略套件來源的 SSL/TLS 憑證檢查，攻擊者可以透過中間人攻擊（MITM）在容器建置期間植入惡意套件，導致整個建置環境被入侵。",
-    "suggestion": "正式環境嚴禁使用 `--no-check-certificate`。若遇到憑證問題，應排查環境配置（如是否缺少根憑證），或是確實解決憑證驗證問題，確保套件來源可信。",
-    "is_new": false
-  },
-  {
-    "level": "warning",
-    "role": "Assassin",
-    "location": "Dockerfile:12",
-    "problem": "使用了 `alpine:latest` 作為基底映像。`latest` 標籤極不穩定，當 Alpine Linux 發布新版本時，建置出的映像環境可能會產生難以預期的變更，甚至引入不相容的套件版本，導致建置再現性（Reproducibility）崩潰。",
-    "suggestion": "建議釘選明確的 Alpine 版本號（例如 `alpine:3.20`），以確保每次建置的執行環境完全一致且可預測。",
-    "is_new": false
-  },
-  {
-    "level": "warning",
-    "role": "Leo",
-    "location": "Dockerfile:21",
-    "problem": "Dockerfile 中留下了明確的安全性警告註解（關於 --no-check-certificate）。這種將安全性風險留在程式碼中並以註解提醒的做法，容易被未來的維護者忽略，導致潛在的安全性漏洞。",
-    "suggestion": "若此 flag 僅為暫時性，應盡快優化環境以移除它；若不可避免，應將此設定集中管理，而非在 Dockerfile 中寫死此類不安全的設定。",
-    "is_new": false
-  },
-  {
-    "level": "info",
-    "role": "Leo",
-    "location": "app/git.js:194",
-    "problem": "`withAskpass` is tightly coupled with Gitea's bot-trigger behavior.",
-    "suggestion": "Encapsulate bot commit logic within `commitAndPush`, auto-selecting the correct `GITEA_COMMENT_TOKEN`.",
-    "is_new": false
-  }
-]
+[]
-- 
2.53.0


From 6c3e7b9d375728701f0a5bbb73a40781e8916103 Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Fri, 26 Jun 2026 09:26:12 +0000
Subject: [PATCH 05/19] =?UTF-8?q?fix(LLM=20=E5=AF=A9=E6=9F=A5=E6=B5=81?=
 =?UTF-8?q?=E7=A8=8B):=20=E9=81=BF=E5=85=8D=E5=96=AE=E4=B8=80=E8=A7=92?=
 =?UTF-8?q?=E8=89=B2=E5=A4=B1=E6=95=97=E4=B8=AD=E6=AD=A2=20pipeline?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 app/llm.js           | 29 +++++++++++++++++++++++------
 app/main.js          | 10 +++++++++-
 app/test/llm.test.js | 14 +++++++++-----
 3 files changed, 41 insertions(+), 12 deletions(-)

diff --git a/app/llm.js b/app/llm.js
index bd63588..71f06cb 100644
--- a/app/llm.js
+++ b/app/llm.js
@@ -1,7 +1,7 @@
 import axios from 'axios';
 import { getLLMConfig, getOpenCodeHttpsAgent } from './config.js';
 import { recordUsage } from './usage.js';
-import { line, error } from './log.js';
+import { line } from './log.js';
 
 /**
  * 將模型識別字串解析為 OpenCode API 所需的 provider 與 model 識別碼。
@@ -49,6 +49,23 @@ function extractOpenCodeContent(data) {
     .join('');
 }
 
+function summarizeErrorResponse(data) {
+  if (data == null) return '';
+  if (typeof data === 'string') return data.slice(0, 500);
+  try {
+    return JSON.stringify(data).slice(0, 500);
+  } catch {
+    return String(data).slice(0, 500);
+  }
+}
+
+function formatOpenCodeError(e) {
+  const status = e.response?.status;
+  const response = summarizeErrorResponse(e.response?.data);
+  const statusText = status ? `HTTP ${status}` : e.message;
+  return response ? `${statusText}: ${response}` : statusText;
+}
+
 /**
  * 對 OpenCode server 執行一次完整對話：建立 session 後送出訊息並回傳結果。
  *
@@ -91,8 +108,8 @@ async function chatOpenCode(baseURL, model, systemPrompt, userContent, headers)
  * 對 OpenCode server 送出一次對話請求並回傳模型純文字回應。
  *
  * 從設定取得 provider/baseURL/model；未設定 provider 時拋錯。成功時記錄
- * usage 並回傳內容。OpenCode 呼叫失敗時會記錄錯誤並以 `process.exit(1)`
- * 終止整個行程（不會回傳）。
+ * usage 並回傳內容。OpenCode 呼叫失敗時會記錄錯誤並向外拋出，讓呼叫端
+ * 決定是否降級、略過單一角色或終止整體流程。
  *
  * @param {string} systemPrompt - 系統提示詞。
  * @param {string} userContent - 使用者輸入內容。
@@ -112,10 +129,10 @@ export async function chat(systemPrompt, userContent) {
     recordUsage(data);
     return content;
   } catch (e) {
-    line(`[LLM] OpenCode 呼叫失敗: ${e.message}`);
+    const message = formatOpenCodeError(e);
+    line(`[LLM] OpenCode 呼叫失敗: ${message}`);
+    throw new Error(message);
   }
-  error('[LLM] OpenCode 呼叫失敗，終止流程');
-  process.exit(1);
 }
 
 /**
diff --git a/app/main.js b/app/main.js
index ec3974e..ca1b967 100644
--- a/app/main.js
+++ b/app/main.js
@@ -122,10 +122,18 @@ async function main() {
   }
   const analyses = await Promise.allSettled(roles.map(role => analyzeWithRole(role, diff)));
   const newFindings = [];
+  let fulfilledAnalyses = 0;
   for (let i = 0; i < analyses.length; i++) {
-    if (analyses[i].status === 'fulfilled') newFindings.push(...analyses[i].value);
+    if (analyses[i].status === 'fulfilled') {
+      fulfilledAnalyses += 1;
+      newFindings.push(...analyses[i].value);
+    }
     else warn(`[${roles[i].name}] 分析失敗（跳過）: ${analyses[i].reason?.message}`);
   }
+  if (fulfilledAnalyses === 0) {
+    result(false, '所有角色分析皆失敗，終止流程以避免誤判為審查通過');
+    process.exit(1);
+  }
   // 對只有檔名、缺行號的問題，反問原角色補上行號（最多重試數次），確保後續能行內標註
   await resolveMissingLineNumbers(newFindings, diff);
   output(`新 findings ${newFindings.length} 筆（${formatFindingsStatsLine(newFindings)}）`);
diff --git a/app/test/llm.test.js b/app/test/llm.test.js
index b99dd94..a23122f 100644
--- a/app/test/llm.test.js
+++ b/app/test/llm.test.js
@@ -85,14 +85,18 @@ describe('chat - OpenCode', async () => {
     assert.equal(result, 'hello world');
   });
 
-  it('calls process.exit(1) when OpenCode fails', async () => {
+  it('throws an error when OpenCode fails instead of exiting the process', async () => {
     process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mock.method(axios, 'post', async () => { throw new Error('fail'); });
-    const exitMock = mock.method(process, 'exit', () => { throw new Error('exit:1'); });
+    mock.method(axios, 'post', async () => {
+      const err = new Error('Request failed with status code 500');
+      err.response = { status: 500, data: { error: 'provider overloaded' } };
+      throw err;
+    });
+    const exitMock = mock.method(process, 'exit', () => { throw new Error('exit should not be called'); });
 
-    await assert.rejects(() => chat('sys', 'user'), /exit:1/);
+    await assert.rejects(() => chat('sys', 'user'), /HTTP 500.*provider overloaded/);
 
-    assert.equal(exitMock.mock.calls[0].arguments[0], 1);
+    assert.equal(exitMock.mock.calls.length, 0);
   });
 });
 
-- 
2.53.0


From 93be261b909299c3fb87163992cab29c6fb57a96 Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Fri, 26 Jun 2026 09:36:33 +0000
Subject: [PATCH 06/19] =?UTF-8?q?fix(OpenCode=20=E5=91=BC=E5=8F=AB):=20?=
 =?UTF-8?q?=E9=99=8D=E4=BD=8E=E4=B8=A6=E8=A1=8C=E8=AB=8B=E6=B1=82=E4=B8=A6?=
 =?UTF-8?q?=E9=87=8D=E8=A9=A6=E6=9A=AB=E6=99=82=E6=80=A7=E9=8C=AF=E8=AA=A4?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 app/llm.js           | 28 +++++++++++++++++++++++++++-
 app/main.js          | 11 ++++++-----
 app/test/llm.test.js | 24 +++++++++++++++++++++++-
 3 files changed, 56 insertions(+), 7 deletions(-)

diff --git a/app/llm.js b/app/llm.js
index 71f06cb..eaf77e4 100644
--- a/app/llm.js
+++ b/app/llm.js
@@ -32,6 +32,10 @@ function opencodeAxiosOptions(headers) {
   };
 }
 
+function sleep(ms) {
+  return new Promise(resolve => setTimeout(resolve, ms));
+}
+
 /**
  * 從 OpenCode 訊息回應中抽取並串接所有文字片段。
  *
@@ -66,6 +70,11 @@ function formatOpenCodeError(e) {
   return response ? `${statusText}: ${response}` : statusText;
 }
 
+function isTransientOpenCodeError(e) {
+  const status = e.response?.status;
+  return status === 500 || status === 502 || status === 503 || status === 504 || status === 429;
+}
+
 /**
  * 對 OpenCode server 執行一次完整對話：建立 session 後送出訊息並回傳結果。
  *
@@ -104,6 +113,23 @@ async function chatOpenCode(baseURL, model, systemPrompt, userContent, headers)
   return { content: extractOpenCodeContent(resp.data), data: resp.data };
 }
 
+async function chatOpenCodeWithRetry(baseURL, model, systemPrompt, userContent, headers) {
+  const maxAttempts = Number(process.env.OPENCODE_RETRY_ATTEMPTS || 3);
+  let lastError;
+  for (let attempt = 1; attempt <= maxAttempts; attempt++) {
+    try {
+      return await chatOpenCode(baseURL, model, systemPrompt, userContent, headers);
+    } catch (e) {
+      lastError = e;
+      if (!isTransientOpenCodeError(e) || attempt === maxAttempts) throw e;
+      const delay = Math.min(1000 * 2 ** (attempt - 1), 8000);
+      line(`[LLM] OpenCode 暫時性錯誤，${delay}ms 後重試 (${attempt}/${maxAttempts}): ${formatOpenCodeError(e)}`);
+      await sleep(delay);
+    }
+  }
+  throw lastError;
+}
+
 /**
  * 對 OpenCode server 送出一次對話請求並回傳模型純文字回應。
  *
@@ -125,7 +151,7 @@ export async function chat(systemPrompt, userContent) {
   const headers = { 'Content-Type': 'application/json' };
 
   try {
-    const { content, data } = await chatOpenCode(baseURL, model, systemPrompt, userContent, headers);
+    const { content, data } = await chatOpenCodeWithRetry(baseURL, model, systemPrompt, userContent, headers);
     recordUsage(data);
     return content;
   } catch (e) {
diff --git a/app/main.js b/app/main.js
index ca1b967..cc7e68f 100644
--- a/app/main.js
+++ b/app/main.js
@@ -120,15 +120,16 @@ async function main() {
   } catch (e) {
     warn(`角色介紹 comment 發布失敗（繼續執行）: ${e.message}`);
   }
-  const analyses = await Promise.allSettled(roles.map(role => analyzeWithRole(role, diff)));
   const newFindings = [];
   let fulfilledAnalyses = 0;
-  for (let i = 0; i < analyses.length; i++) {
-    if (analyses[i].status === 'fulfilled') {
+  for (const role of roles) {
+    try {
+      const findings = await analyzeWithRole(role, diff);
       fulfilledAnalyses += 1;
-      newFindings.push(...analyses[i].value);
+      newFindings.push(...findings);
+    } catch (e) {
+      warn(`[${role.name}] 分析失敗（跳過）: ${e.message}`);
     }
-    else warn(`[${roles[i].name}] 分析失敗（跳過）: ${analyses[i].reason?.message}`);
   }
   if (fulfilledAnalyses === 0) {
     result(false, '所有角色分析皆失敗，終止流程以避免誤判為審查通過');
diff --git a/app/test/llm.test.js b/app/test/llm.test.js
index a23122f..ffb3e1e 100644
--- a/app/test/llm.test.js
+++ b/app/test/llm.test.js
@@ -4,7 +4,7 @@ import axios from 'axios';
 import { extractBalancedJSON, extractJSONText } from '../llm.js';
 
 const ENV_KEYS = [
-  'OPENCODE_BASE_URL', 'OPENCODE_MODEL', 'OPENCODE_PROVIDER',
+  'OPENCODE_BASE_URL', 'OPENCODE_MODEL', 'OPENCODE_PROVIDER', 'OPENCODE_RETRY_ATTEMPTS',
 ];
 
 let saved = {};
@@ -87,6 +87,7 @@ describe('chat - OpenCode', async () => {
 
   it('throws an error when OpenCode fails instead of exiting the process', async () => {
     process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
+    process.env.OPENCODE_RETRY_ATTEMPTS = '1';
     mock.method(axios, 'post', async () => {
       const err = new Error('Request failed with status code 500');
       err.response = { status: 500, data: { error: 'provider overloaded' } };
@@ -98,6 +99,27 @@ describe('chat - OpenCode', async () => {
 
     assert.equal(exitMock.mock.calls.length, 0);
   });
+
+  it('retries transient OpenCode failures before returning content', async () => {
+    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
+    process.env.OPENCODE_RETRY_ATTEMPTS = '2';
+    let messageAttempts = 0;
+    mock.method(axios, 'post', async (url) => {
+      if (url.endsWith('/session')) return { data: { id: 'ses_test' } };
+      messageAttempts += 1;
+      if (messageAttempts === 1) {
+        const err = new Error('Request failed with status code 500');
+        err.response = { status: 500, data: { error: 'temporary failure' } };
+        throw err;
+      }
+      return { data: { parts: [{ type: 'text', text: 'ok after retry' }] } };
+    });
+
+    const result = await chat('sys', 'user');
+
+    assert.equal(result, 'ok after retry');
+    assert.equal(messageAttempts, 2);
+  });
 });
 
 describe('chatJSON', async () => {
-- 
2.53.0


From 08a72a8c7fda46e14da2d92b711b714a8725e8fd Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Sat, 27 Jun 2026 13:30:40 +0000
Subject: [PATCH 07/19] =?UTF-8?q?feat(ai-code-review):=20=E6=94=B9?=
 =?UTF-8?q?=E7=94=A8=20AI=20=E5=8A=A9=E7=90=86=20CLI=20=E5=9F=B7=E8=A1=8C?=
 =?UTF-8?q?=E5=AF=A9=E6=9F=A5?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 action.yaml      |  49 ++++++-----
 app/config.js    |  62 +++++++++----
 app/llm.js       | 225 ++++++++++++++++++++---------------------------
 app/preflight.js |  64 +++-----------
 4 files changed, 177 insertions(+), 223 deletions(-)

diff --git a/action.yaml b/action.yaml
index 31e568a..d8f86a7 100644
--- a/action.yaml
+++ b/action.yaml
@@ -8,27 +8,30 @@ inputs:
   comment_token:
     description: ''
     required: false
-  opencode_base_url:
-    description: 'OpenCode server Base URL'
-    required: false
-  opencode_model:
-    description: 'OpenCode model id'
-    required: false
-  opencode_provider:
-    description: 'OpenCode server provider id'
-    required: false
+  model:
+    description: ''
+    required: true
 runs:
-  using: 'docker'
-  image: 'Dockerfile'
-  env:
-    GITEA_SERVER_URL: ${{ gitea.server_url }}
-    GITEA_REPOSITORY: ${{ gitea.repository }}
-    GITEA_TOKEN: ${{ inputs.token || gitea.token }}
-    GITEA_COMMENT_TOKEN: ${{ inputs.comment_token || inputs.token || gitea.token }}
-    PR_NUMBER: ${{ gitea.event.pull_request.number }}
-    PR_HEAD_SHA: ${{ gitea.event.pull_request.head.sha }}
-    PR_HEAD_BRANCH: ${{ gitea.event.pull_request.head.ref }}
-    PR_BASE_BRANCH: ${{ gitea.event.pull_request.base.ref }}
-    OPENCODE_BASE_URL: ${{ inputs.opencode_base_url }}
-    OPENCODE_MODEL: ${{ inputs.opencode_model }}
-    OPENCODE_PROVIDER: ${{ inputs.opencode_provider }}
+  using: 'composite'
+  steps:
+  - name: 安裝 Node.js 相依套件
+    shell: bash
+    run: |
+      cd "${GITHUB_ACTION_PATH:-${GITEA_ACTION_PATH:-.}}/app"
+      npm ci
+  - name: 執行 AI 程式碼審查
+    shell: bash
+    env:
+      GITEA_SERVER_URL: ${{ gitea.server_url }}
+      GITEA_REPOSITORY: ${{ gitea.repository }}
+      GITEA_TOKEN: ${{ inputs.token || gitea.token }}
+      GITEA_COMMENT_TOKEN: ${{ inputs.comment_token || inputs.token || gitea.token }}
+      PR_NUMBER: ${{ gitea.event.pull_request.number }}
+      PR_HEAD_SHA: ${{ gitea.event.pull_request.head.sha }}
+      PR_HEAD_BRANCH: ${{ gitea.event.pull_request.head.ref }}
+      PR_BASE_BRANCH: ${{ gitea.event.pull_request.base.ref }}
+      MODEL: ${{ inputs.model }}
+    run: |
+      echo "🚀 AI Code Review Action 啟動"
+      cd "${GITHUB_ACTION_PATH:-${GITEA_ACTION_PATH:-.}}/app"
+      node main.js
diff --git a/app/config.js b/app/config.js
index c92f2b1..e76389c 100644
--- a/app/config.js
+++ b/app/config.js
@@ -1,4 +1,5 @@
 import https from 'https';
+import { execFileSync } from 'child_process';
 
 process.env.NODE_TLS_REJECT_UNAUTHORIZED = '0';
 
@@ -30,25 +31,56 @@ export function getInsecureHttpsAgent() {
 
 export const getOpenCodeHttpsAgent = getInsecureHttpsAgent;
 
+const CLI_CANDIDATES = [
+  {
+    provider: 'codex',
+    command: 'codex',
+    defaultModel: 'gpt-5',
+  },
+  {
+    provider: 'claude',
+    command: 'claude',
+    defaultModel: 'sonnet',
+  },
+  {
+    provider: 'opencode',
+    command: 'opencode',
+    defaultModel: 'google/gemini-2.5-flash',
+  },
+];
+
+function commandExists(command) {
+  try {
+    execFileSync('/bin/sh', ['-lc', `command -v ${command}`], { stdio: 'ignore' });
+    return true;
+  } catch {
+    return false;
+  }
+}
+
 /**
  * 依環境變數解析並回傳 LLM 提供者設定。
  *
- * 當設定了 `OPENCODE_BASE_URL` 時回傳 OpenCode 提供者設定
- * （model 取自 `OPENCODE_MODEL`，預設為 `gemini-2.5-flash`）；
- * 否則回傳各欄位皆為空／null 的「無提供者」設定，由呼叫端據此判斷是否略過 LLM 流程。
+ * 優先使用 `AI_ASSISTANT_CLI` 指定的 CLI；未指定時依序偵測 codex、claude、opencode。
+ * model 優先取 `MODEL`，再相容舊的 `OPENCODE_MODEL`，最後使用各 CLI 預設值。
  *
- * @remarks 每次呼叫都會即時讀取 `process.env`。`apiKeys` 在 OpenCode 模式下為固定佔位值 `['opencode']`，並非真實金鑰。
- * @returns {{ provider: ('opencode'|null), apiKeys: string[], baseURL: (string|null), model: (string|null) }}
+ * @param {{ commandExistsFn?: (command: string) => boolean }} [deps] - 可注入的 CLI 偵測函式，供測試使用。
+ * @returns {{ provider: ('codex'|'claude'|'opencode'|null), apiKeys: string[], baseURL: null, model: (string|null), command: (string|null) }}
  *          LLM 設定物件；`provider` 為 `null` 表示沒有可用的提供者。
  */
-export function getLLMConfig() {
-  if (process.env.OPENCODE_BASE_URL) {
-    return {
-      provider: 'opencode',
-      apiKeys: ['opencode'],
-      baseURL: process.env.OPENCODE_BASE_URL,
-      model: process.env.OPENCODE_MODEL || 'gemini-2.5-flash',
-    };
-  }
-  return { provider: null, apiKeys: [], baseURL: null, model: null };
+export function getLLMConfig({ commandExistsFn = commandExists } = {}) {
+  const requested = process.env.AI_ASSISTANT_CLI;
+  const candidates = requested
+    ? CLI_CANDIDATES.filter(c => c.provider === requested || c.command === requested)
+    : CLI_CANDIDATES;
+  const cli = candidates.find(c => commandExistsFn(c.command));
+  if (!cli) return { provider: null, apiKeys: [], baseURL: null, model: null, command: null };
+
+  return {
+    provider: cli.provider,
+    apiKeys: [cli.provider],
+    baseURL: null,
+    model: process.env.MODEL || process.env.OPENCODE_MODEL || cli.defaultModel,
+    command: cli.command,
+  };
 }
diff --git a/app/llm.js b/app/llm.js
index eaf77e4..7a27866 100644
--- a/app/llm.js
+++ b/app/llm.js
@@ -1,168 +1,129 @@
-import axios from 'axios';
-import { getLLMConfig, getOpenCodeHttpsAgent } from './config.js';
+import * as childProcess from 'child_process';
+import { mkdtemp, writeFile, rm } from 'fs/promises';
+import { tmpdir } from 'os';
+import { join } from 'path';
+import { getLLMConfig } from './config.js';
 import { recordUsage } from './usage.js';
 import { line } from './log.js';
 
 /**
- * 將模型識別字串解析為 OpenCode API 所需的 provider 與 model 識別碼。
- *
- * 當字串含有 `/` 時視為 `providerID/modelID` 形式並拆解；否則 provider
- * 取環境變數 `OPENCODE_PROVIDER`（預設 `google`），model 則為整個字串。
- *
- * @param {string} model - 模型識別字串，例如 `"google/gemini-2.0"` 或 `"gemini-2.0"`。
- * @returns {{ providerID: string, modelID: string }} 拆解後的 provider 與 model 識別碼。
+ * 將既有 system/user prompt 合併成一次 CLI 呼叫用的輸入。
  */
-function opencodeModelConfig(model) {
-  const [providerID, modelID] = model.includes('/') ? model.split('/', 2) : [process.env.OPENCODE_PROVIDER || 'google', model];
-  return { providerID, modelID };
+function buildPrompt(systemPrompt, userContent) {
+  return [
+    '請依照以下系統指示處理使用者內容，並只輸出要求的最終結果。',
+    '',
+    '<system>',
+    systemPrompt,
+    '</system>',
+    '',
+    '<user>',
+    userContent,
+    '</user>',
+  ].join('\n');
 }
 
-/**
- * 建立傳給 axios 的共用請求選項，統一注入 headers 與 OpenCode 專用的 HTTPS agent。
- *
- * 供本模組所有 OpenCode HTTP 呼叫共用，集中管理連線設定。
- *
- * @param {Record<string, string>} headers - 要附加於請求的 HTTP 標頭。
- * @returns {{ headers: Record<string, string>, httpsAgent: import('https').Agent }} axios 請求選項物件。
- */
-function opencodeAxiosOptions(headers) {
-  return {
-    headers,
-    httpsAgent: getOpenCodeHttpsAgent(),
-  };
+function cliArgs(provider, model, promptFile = null) {
+  if (provider === 'codex') {
+    return ['exec', '--model', model, '--sandbox', 'read-only', '--ask-for-approval', 'never', '--skip-git-repo-check', '-'];
+  }
+  if (provider === 'claude') {
+    return ['--print', '--model', model, '--permission-mode', 'dontAsk', '--no-session-persistence'];
+  }
+  if (provider === 'opencode') {
+    return ['run', '--model', model, '--format', 'default', '--file', promptFile, '請依附件 prompt.md 的完整內容執行，並只輸出要求的最終結果。'];
+  }
+  throw new Error(`不支援的 AI 助理 CLI: ${provider}`);
 }
 
-function sleep(ms) {
-  return new Promise(resolve => setTimeout(resolve, ms));
+function summarizeCliError(e) {
+  const stderr = String(e.stderr || '').trim();
+  const stdout = String(e.stdout || '').trim();
+  return (stderr || stdout || e.message || String(e)).slice(0, 1000);
 }
 
-/**
- * 從 OpenCode 訊息回應中抽取並串接所有文字片段。
- *
- * 以多重 fallback 相容不同包裹層級的回應結構（`parts` / `data.parts` /
- * `info.content` / `data.info.content`），逐片段取 `text` 或 `content` 後串接。
- *
- * @param {object} data - OpenCode `/session/{id}/message` 的回應資料物件。
- * @returns {string} 串接後的純文字內容；無可用片段時回傳空字串。
- */
-function extractOpenCodeContent(data) {
-  const parts = data.parts || data.data?.parts || data.info?.content || data.data?.info?.content || [];
-  return parts
-    .map(part => part.text || part.content || '')
-    .filter(Boolean)
-    .join('');
-}
-
-function summarizeErrorResponse(data) {
-  if (data == null) return '';
-  if (typeof data === 'string') return data.slice(0, 500);
+async function runAssistantCLI({ provider, command, model }, prompt) {
+  let tempDir = null;
+  let promptFile = null;
+  if (provider === 'opencode') {
+    tempDir = await mkdtemp(join(tmpdir(), 'ai-review-prompt-'));
+    promptFile = join(tempDir, 'prompt.md');
+    await writeFile(promptFile, prompt);
+  }
+  const args = cliArgs(provider, model, promptFile);
+  const maxBuffer = Number(process.env.AI_ASSISTANT_MAX_BUFFER || 20 * 1024 * 1024);
+  const timeout = Number(process.env.AI_ASSISTANT_TIMEOUT_MS || 15 * 60 * 1000);
   try {
-    return JSON.stringify(data).slice(0, 500);
-  } catch {
-    return String(data).slice(0, 500);
+    return await new Promise((resolve, reject) => {
+      const child = childProcess.spawn(command, args, { env: process.env, stdio: ['pipe', 'pipe', 'pipe'] });
+      let stdout = '';
+      let stderr = '';
+      let settled = false;
+      const timer = setTimeout(() => {
+        settled = true;
+        child.kill('SIGTERM');
+        reject(new Error(`${provider} CLI 逾時 (${timeout}ms)`));
+      }, timeout);
+
+      const append = (kind, chunk) => {
+        if (kind === 'stdout') stdout += chunk;
+        else stderr += chunk;
+        if (stdout.length + stderr.length > maxBuffer) {
+          settled = true;
+          child.kill('SIGTERM');
+          reject(new Error(`${provider} CLI 輸出超過 ${maxBuffer} bytes`));
+        }
+      };
+
+      child.stdout.setEncoding('utf8');
+      child.stderr.setEncoding('utf8');
+      child.stdout.on('data', chunk => append('stdout', chunk));
+      child.stderr.on('data', chunk => append('stderr', chunk));
+      child.on('error', reject);
+      child.on('close', (code, signal) => {
+        clearTimeout(timer);
+        if (settled) return;
+        if (code === 0) resolve(stdout.trim());
+        else reject(Object.assign(new Error(`${provider} CLI exited with ${code ?? signal}`), { stdout, stderr }));
+      });
+      child.stdin.end(provider === 'opencode' ? '' : prompt);
+    });
+  } finally {
+    if (tempDir) await rm(tempDir, { recursive: true, force: true });
   }
 }
 
-function formatOpenCodeError(e) {
-  const status = e.response?.status;
-  const response = summarizeErrorResponse(e.response?.data);
-  const statusText = status ? `HTTP ${status}` : e.message;
-  return response ? `${statusText}: ${response}` : statusText;
-}
-
-function isTransientOpenCodeError(e) {
-  const status = e.response?.status;
-  return status === 500 || status === 502 || status === 503 || status === 504 || status === 429;
-}
-
 /**
- * 對 OpenCode server 執行一次完整對話：建立 session 後送出訊息並回傳結果。
+ * 對目前環境可用的 AI 助理 CLI 送出一次對話請求並回傳純文字回應。
  *
- * 先 POST `/session` 取得 session id（缺少則拋錯），再 POST
- * `/session/{id}/message` 送出 system prompt 與使用者內容，最後抽取回應文字。
- * 會發出兩次 HTTP 請求；網路或 API 錯誤會向外拋出，交由呼叫端處理。
- *
- * @param {string} baseURL - OpenCode server 基底 URL（尾端斜線會被去除）。
- * @param {string} model - 模型識別字串，將交由 {@link opencodeModelConfig} 解析。
- * @param {string} systemPrompt - 系統提示詞。
- * @param {string} userContent - 使用者輸入內容。
- * @param {Record<string, string>} headers - 附加於請求的 HTTP 標頭。
- * @returns {Promise<{ content: string, data: object }>} 抽取後的文字內容與原始回應資料。
- * @throws {Error} 當回應中無 session id，或任一 HTTP 請求失敗時。
- */
-async function chatOpenCode(baseURL, model, systemPrompt, userContent, headers) {
-  const base = baseURL.replace(/\/$/, '');
-  const { providerID, modelID } = opencodeModelConfig(model);
-  const session = await axios.post(
-    `${base}/session`,
-    { title: 'AI Code Review', model: { providerID, id: modelID } },
-    opencodeAxiosOptions(headers)
-  );
-  const sessionID = session.data.id || session.data.data?.id;
-  if (!sessionID) throw new Error('OpenCode session 建立失敗：回應中沒有 session id');
-
-  const resp = await axios.post(
-    `${base}/session/${sessionID}/message`,
-    {
-      model: { providerID, modelID },
-      system: systemPrompt,
-      parts: [{ type: 'text', text: userContent }],
-    },
-    opencodeAxiosOptions(headers)
-  );
-  return { content: extractOpenCodeContent(resp.data), data: resp.data };
-}
-
-async function chatOpenCodeWithRetry(baseURL, model, systemPrompt, userContent, headers) {
-  const maxAttempts = Number(process.env.OPENCODE_RETRY_ATTEMPTS || 3);
-  let lastError;
-  for (let attempt = 1; attempt <= maxAttempts; attempt++) {
-    try {
-      return await chatOpenCode(baseURL, model, systemPrompt, userContent, headers);
-    } catch (e) {
-      lastError = e;
-      if (!isTransientOpenCodeError(e) || attempt === maxAttempts) throw e;
-      const delay = Math.min(1000 * 2 ** (attempt - 1), 8000);
-      line(`[LLM] OpenCode 暫時性錯誤，${delay}ms 後重試 (${attempt}/${maxAttempts}): ${formatOpenCodeError(e)}`);
-      await sleep(delay);
-    }
-  }
-  throw lastError;
-}
-
-/**
- * 對 OpenCode server 送出一次對話請求並回傳模型純文字回應。
- *
- * 從設定取得 provider/baseURL/model；未設定 provider 時拋錯。成功時記錄
- * usage 並回傳內容。OpenCode 呼叫失敗時會記錄錯誤並向外拋出，讓呼叫端
- * 決定是否降級、略過單一角色或終止整體流程。
+ * 從設定取得 provider/command/model；未偵測到 CLI 時拋錯。成功時記錄一次
+ * usage 呼叫（CLI 通常不回傳 token 明細，因此 token 可能為 0）並回傳內容。
  *
  * @param {string} systemPrompt - 系統提示詞。
  * @param {string} userContent - 使用者輸入內容。
  * @returns {Promise<string>} 模型回應的純文字內容。
- * @throws {Error} 當未設定 OpenCode server（缺少 provider）時。
+ * @throws {Error} 當未偵測到可用 AI 助理 CLI，或 CLI 呼叫失敗時。
  */
 export async function chat(systemPrompt, userContent) {
-  const { provider, baseURL, model } = getLLMConfig();
-  if (!provider) throw new Error('未設定 OpenCode server，請設定 OPENCODE_BASE_URL');
+  const cfg = getLLMConfig();
+  const { provider, command, model } = cfg;
+  if (!provider || !command) throw new Error('未偵測到可用 AI 助理 CLI，請安裝 codex、claude 或 opencode');
 
-  line(`[LLM] provider=${provider} model=${model}`);
-
-  const headers = { 'Content-Type': 'application/json' };
+  line(`[LLM] provider=${provider} command=${command} model=${model}`);
 
   try {
-    const { content, data } = await chatOpenCodeWithRetry(baseURL, model, systemPrompt, userContent, headers);
-    recordUsage(data);
+    const content = await runAssistantCLI(cfg, buildPrompt(systemPrompt, userContent));
+    recordUsage(null);
     return content;
   } catch (e) {
-    const message = formatOpenCodeError(e);
-    line(`[LLM] OpenCode 呼叫失敗: ${message}`);
+    const message = summarizeCliError(e);
+    line(`[LLM] ${provider} CLI 呼叫失敗: ${message}`);
     throw new Error(message);
   }
 }
 
 /**
- * 對 OpenCode 送出對話並將回應解析為 JSON 物件/陣列。
+ * 對 AI 助理 CLI 送出對話並將回應解析為 JSON 物件/陣列。
  *
  * 先取得文字回應，經 {@link extractJSONText} 抽出 JSON 片段後解析。
  * 解析失敗時記錄錯誤並回傳空陣列，不向外拋錯（容錯設計）。
diff --git a/app/preflight.js b/app/preflight.js
index 90b8814..ca7586d 100644
--- a/app/preflight.js
+++ b/app/preflight.js
@@ -6,7 +6,6 @@ import {
   GITEA_REPOSITORY,
   PR_NUMBER,
   getInsecureHttpsAgent,
-  getOpenCodeHttpsAgent,
   getLLMConfig,
 } from './config.js';
 import { verifyRemoteAccess } from './git.js';
@@ -30,33 +29,6 @@ const api = (path) => `${GITEA_SERVER_URL.replace(/\/$/, '')}/api/v1${path}`;
  * @returns {{Authorization: string, 'Content-Type': string}} 可直接交給 axios 的 headers 物件。
  */
 const giteaHeaders = (token) => ({ Authorization: `token ${token}`, 'Content-Type': 'application/json' });
-/**
- * 將模型字串解析為 OpenCode 的 providerID 與 modelID。
- *
- * 若 model 含 `/` 則以斜線拆分為 provider/model；否則 provider 取
- * 環境變數 OPENCODE_PROVIDER（預設 `google`），model 即原字串。
- * @param {string} model - 模型識別字串，例如 `google/gemini-2.0` 或 `gemini-2.0`。
- * @returns {{providerID: string, modelID: string}} 解析後的 provider 與 model 識別碼。
- * @remarks 讀取 process.env.OPENCODE_PROVIDER；split 上限為 2 段，多餘段落會被忽略。
- */
-const opencodeModelConfig = (model) => {
-  const [providerID, modelID] = model.includes('/') ? model.split('/', 2) : [process.env.OPENCODE_PROVIDER || 'google', model];
-  return { providerID, modelID };
-};
-/**
- * 組出呼叫 OpenCode server 用的 axios 請求設定。
- *
- * 固定 30 秒逾時，並套用 config.js 的 getOpenCodeHttpsAgent() 作為 httpsAgent。
- * @param {object} headers - 要套用的 HTTP headers 物件。
- * @returns {{headers: object, timeout: number, httpsAgent: import('https').Agent}} axios 設定物件。
- * @remarks 每次呼叫都會執行 getOpenCodeHttpsAgent() 取得 agent。
- */
-const opencodeAxiosOptions = (headers) => ({
-  headers,
-  timeout: 30000,
-  httpsAgent: getOpenCodeHttpsAgent(),
-});
-
 /**
  * 將（axios）錯誤格式化為易讀的訊息字串。
  *
@@ -124,37 +96,23 @@ export async function verifyCommentToken(token = GITEA_COMMENT_TOKEN) {
 }
 
 /**
- * 驗證 LLM（OpenCode server）設定可用。
+ * 驗證 LLM（AI 助理 CLI）設定可用。
  *
- * 依序確認：已設定 provider、有 base URL、health 端點可連線、OpenCode 已設定
- * 對應 provider 且其列出指定 model。任一不符回傳對應錯誤；錯誤被攔截不會 throw。
+ * 確認目前環境可偵測到支援的 CLI，且已解析出 model。實際模型可用性由 CLI
+ * 在正式呼叫時回報；preflight 不主動送 prompt，避免額外消耗額度。
  * @returns {Promise<{ok: true, provider: string}|{ok: false, provider?: string, error: string}>}
  *   通過時含 provider；未設定 provider 的失敗分支不含 provider 欄位。
- * @remarks 設定來源為 config.js 的 getLLMConfig()；provider/model 鍵的比對使用 opencodeModelConfig 解析後的 providerID/modelID。
+ * @remarks 設定來源為 config.js 的 getLLMConfig()。
  */
 export async function verifyLLM() {
-  const { provider, baseURL, model } = getLLMConfig();
-  if (!provider) return { ok: false, error: '未設定 OpenCode server，請設定 OPENCODE_BASE_URL' };
-  if (!baseURL) return { ok: false, provider, error: `${provider} 缺少 base URL` };
-
-  const base = baseURL.replace(/\/$/, '');
-  const headers = { 'Content-Type': 'application/json' };
-
-  const { providerID, modelID } = opencodeModelConfig(model);
-  try {
-    await axios.get(`${base}/global/health`, opencodeAxiosOptions(headers));
-    const providers = await axios.get(`${base}/config/providers`, opencodeAxiosOptions(headers));
-    const configuredProvider = providers.data.providers?.find(p => p.id === providerID);
-    if (!configuredProvider) return { ok: false, provider, error: `OpenCode server 未設定 provider=${providerID}` };
-    if (!configuredProvider.models?.[modelID]) return { ok: false, provider, error: `OpenCode server provider=${providerID} 未列出 model=${modelID}` };
-    return { ok: true, provider };
-  } catch (e) {
-    return { ok: false, provider, error: `OpenCode server 驗證失敗: ${e.message}` };
-  }
+  const { provider, command, model } = getLLMConfig();
+  if (!provider || !command) return { ok: false, error: '未偵測到可用 AI 助理 CLI，請安裝 codex、claude 或 opencode' };
+  if (!model) return { ok: false, provider, error: '未設定 MODEL' };
+  return { ok: true, provider, command, model };
 }
 
 /**
- * 執行所有前置驗證（Step2）：環境變數、Gitea token、comment token、git 遠端、LLM。
+ * 執行所有前置驗證（Step2）：環境變數、Gitea token、comment token、git 遠端、LLM CLI。
  *
  * 全程唯讀，不發布任何 comment；任一檢查失敗即記錄錯誤並回傳 false。
  * 各檢查可經 deps 注入覆寫，方便單元測試。
@@ -164,7 +122,7 @@ export async function verifyLLM() {
  * @param {Function} [deps.verifyToken=verifyGiteaToken] - Gitea token / repo 讀取驗證。
  * @param {Function} [deps.verifyComment=verifyCommentToken] - comment token 驗證。
  * @param {Function} [deps.verifyRemote=verifyRemoteAccess] - git 遠端（ls-remote）認證驗證。
- * @param {Function} [deps.verifyLLMFn=verifyLLM] - LLM（OpenCode）連線驗證。
+ * @param {Function} [deps.verifyLLMFn=verifyLLM] - LLM（AI 助理 CLI）驗證。
  * @returns {Promise<boolean>} 全部通過為 true，任一失敗為 false。
  * @remarks 透過 log.js 輸出 step/ok/line/error/result 記錄；不會 throw（前提是注入的檢查函式皆自行攔截錯誤）。
  */
@@ -212,7 +170,7 @@ export async function runPreflight(workspace = process.env.GITHUB_WORKSPACE || '
     error(`LLM 驗證失敗: ${llm.error}`);
     return false;
   }
-  ok(`LLM provider=${llm.provider} 連線正常`);
+  ok(`LLM provider=${llm.provider} CLI 可用`);
 
   result(true, '前置驗證通過');
   return true;
-- 
2.53.0


From 4fcb240208d5003d2e708fc90a0b062d0859df83 Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Sat, 27 Jun 2026 13:30:45 +0000
Subject: [PATCH 08/19] =?UTF-8?q?test(ai-code-review):=20=E6=9B=B4?=
 =?UTF-8?q?=E6=96=B0=20CLI=20=E5=91=BC=E5=8F=AB=E8=88=87=E5=89=8D=E7=BD=AE?=
 =?UTF-8?q?=E9=A9=97=E8=AD=89=E6=B8=AC=E8=A9=A6?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 app/test/config.test.js    |  35 +++++----
 app/test/llm.test.js       | 151 ++++++++++++++-----------------------
 app/test/preflight.test.js |  96 +++++++++--------------
 3 files changed, 115 insertions(+), 167 deletions(-)

diff --git a/app/test/config.test.js b/app/test/config.test.js
index 7139dae..942acc3 100644
--- a/app/test/config.test.js
+++ b/app/test/config.test.js
@@ -3,7 +3,7 @@ import assert from 'node:assert/strict';
 import { getLLMConfig, getOpenCodeHttpsAgent } from '../config.js';
 
 const ENV_KEYS = [
-  'OPENCODE_BASE_URL', 'OPENCODE_MODEL', 'OPENCODE_PROVIDER',
+  'AI_ASSISTANT_CLI', 'MODEL', 'OPENCODE_MODEL',
 ];
 
 let saved = {};
@@ -20,26 +20,35 @@ afterEach(() => {
 
 describe('getLLMConfig', () => {
   it('returns null provider when no env vars set', () => {
-    const cfg = getLLMConfig();
+    const cfg = getLLMConfig({ commandExistsFn: () => false });
     assert.equal(cfg.provider, null);
     assert.deepEqual(cfg.apiKeys, []);
+    assert.equal(cfg.command, null);
   });
 
-  it('detects opencode server with defaults', () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    const cfg = getLLMConfig();
-    assert.equal(cfg.provider, 'opencode');
-    assert.deepEqual(cfg.apiKeys, ['opencode']);
-    assert.equal(cfg.baseURL, 'http://opencode.local:4096');
-    assert.equal(cfg.model, 'gemini-2.5-flash');
+  it('detects the first installed assistant CLI with defaults', () => {
+    const cfg = getLLMConfig({ commandExistsFn: command => command === 'claude' });
+    assert.equal(cfg.provider, 'claude');
+    assert.deepEqual(cfg.apiKeys, ['claude']);
+    assert.equal(cfg.baseURL, null);
+    assert.equal(cfg.command, 'claude');
+    assert.equal(cfg.model, 'sonnet');
   });
 
-  it('detects opencode server with custom model', () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
+  it('uses MODEL for the selected assistant CLI', () => {
+    process.env.MODEL = 'gpt-5-mini';
+    const cfg = getLLMConfig({ commandExistsFn: command => command === 'codex' });
+    assert.equal(cfg.provider, 'codex');
+    assert.equal(cfg.command, 'codex');
+    assert.equal(cfg.model, 'gpt-5-mini');
+  });
+
+  it('can force a CLI with AI_ASSISTANT_CLI', () => {
+    process.env.AI_ASSISTANT_CLI = 'opencode';
     process.env.OPENCODE_MODEL = 'google/gemini-2.5-pro';
-    const cfg = getLLMConfig();
+    const cfg = getLLMConfig({ commandExistsFn: command => command === 'codex' || command === 'opencode' });
     assert.equal(cfg.provider, 'opencode');
-    assert.equal(cfg.baseURL, 'http://opencode.local:4096');
+    assert.equal(cfg.command, 'opencode');
     assert.equal(cfg.model, 'google/gemini-2.5-pro');
   });
 
diff --git a/app/test/llm.test.js b/app/test/llm.test.js
index ffb3e1e..5d7bd5c 100644
--- a/app/test/llm.test.js
+++ b/app/test/llm.test.js
@@ -1,133 +1,96 @@
 import { describe, it, beforeEach, afterEach, mock } from 'node:test';
 import assert from 'node:assert/strict';
-import axios from 'axios';
+import { mkdtemp, writeFile, chmod, rm, readFile } from 'fs/promises';
+import { tmpdir } from 'os';
+import { join } from 'path';
 import { extractBalancedJSON, extractJSONText } from '../llm.js';
 
 const ENV_KEYS = [
-  'OPENCODE_BASE_URL', 'OPENCODE_MODEL', 'OPENCODE_PROVIDER', 'OPENCODE_RETRY_ATTEMPTS',
+  'AI_ASSISTANT_CLI', 'MODEL', 'OPENCODE_MODEL', 'PATH', 'AI_ASSISTANT_TIMEOUT_MS', 'AI_ASSISTANT_MAX_BUFFER',
+  'FAKE_AI_STDOUT', 'FAKE_AI_STDERR', 'FAKE_AI_EXIT', 'FAKE_AI_STDIN_PATH', 'FAKE_AI_ARGS_PATH',
 ];
 
 let saved = {};
+let tempDir;
 beforeEach(() => {
   saved = {};
   for (const k of ENV_KEYS) { saved[k] = process.env[k]; delete process.env[k]; }
+  tempDir = null;
 });
-afterEach(() => {
+afterEach(async () => {
   for (const k of ENV_KEYS) {
     if (saved[k] === undefined) delete process.env[k];
     else process.env[k] = saved[k];
   }
+  if (tempDir) await rm(tempDir, { recursive: true, force: true });
   mock.restoreAll();
 });
 
-function mockOpenCodeResponse(content) {
-  let calls = 0;
-  mock.method(axios, 'post', async () => {
-    calls += 1;
-    if (calls === 1) return { data: { id: 'ses_test' } };
-    return { data: { parts: [{ type: 'text', text: content }] } };
-  });
+async function installFakeCLI(command = 'codex') {
+  tempDir = await mkdtemp(join(tmpdir(), 'ai-cli-test-'));
+  const script = join(tempDir, command);
+  await writeFile(script, `#!/bin/sh
+if [ -n "$FAKE_AI_ARGS_PATH" ]; then printf '%s\\n' "$*" > "$FAKE_AI_ARGS_PATH"; fi
+if [ -n "$FAKE_AI_STDIN_PATH" ]; then cat > "$FAKE_AI_STDIN_PATH"; else cat >/dev/null; fi
+if [ -n "$FAKE_AI_STDERR" ]; then printf '%s' "$FAKE_AI_STDERR" >&2; fi
+if [ -n "$FAKE_AI_STDOUT" ]; then printf '%s' "$FAKE_AI_STDOUT"; fi
+exit "\${FAKE_AI_EXIT:-0}"
+`);
+  await chmod(script, 0o755);
+  process.env.PATH = `${tempDir}:${saved.PATH || ''}`;
+  return { stdinPath: join(tempDir, 'stdin.txt'), argsPath: join(tempDir, 'args.txt') };
 }
 
-describe('chat - OpenCode', async () => {
+describe('chat - assistant CLI', async () => {
   const { chat } = await import('../llm.js');
 
-  it('uses OpenCode server session API', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_PROVIDER = 'google';
-    process.env.OPENCODE_MODEL = 'gemini-2.5-flash';
-    const calls = [];
-    mock.method(axios, 'post', async (url, payload, opts) => {
-      calls.push({ url, payload, headers: opts.headers });
-      if (url.endsWith('/session')) return { data: { id: 'ses_test' } };
-      return { data: { parts: [{ type: 'text', text: 'opencode response' }] } };
-    });
+  it('runs the detected CLI with MODEL and sends the prompts through stdin', async () => {
+    const { stdinPath, argsPath } = await installFakeCLI('codex');
+    process.env.MODEL = 'gpt-5-mini';
+    process.env.FAKE_AI_STDOUT = 'cli response';
+    process.env.FAKE_AI_STDIN_PATH = stdinPath;
+    process.env.FAKE_AI_ARGS_PATH = argsPath;
 
     const result = await chat('sys', 'user');
 
-    assert.equal(result, 'opencode response');
-    assert.equal(calls[0].url, 'http://opencode.local:4096/session');
-    assert.deepEqual(calls[0].payload.model, { providerID: 'google', id: 'gemini-2.5-flash' });
-    assert.equal(calls[1].url, 'http://opencode.local:4096/session/ses_test/message');
-    assert.deepEqual(calls[1].payload.model, { providerID: 'google', modelID: 'gemini-2.5-flash' });
-    assert.equal(calls[1].payload.system, 'sys');
-    assert.deepEqual(calls[1].payload.parts, [{ type: 'text', text: 'user' }]);
-    assert.equal(calls[1].headers['Authorization'], undefined);
+    assert.equal(result, 'cli response');
+    assert.match(await readFile(argsPath, 'utf8'), /exec --model gpt-5-mini/);
+    const prompt = await readFile(stdinPath, 'utf8');
+    assert.match(prompt, /<system>\nsys\n<\/system>/);
+    assert.match(prompt, /<user>\nuser\n<\/user>/);
   });
 
-  it('passes an insecure https agent to OpenCode by default', async () => {
-    process.env.OPENCODE_BASE_URL = 'https://opencode.local:4096';
-    const agents = [];
-    mock.method(axios, 'post', async (url, _payload, opts) => {
-      agents.push(opts.httpsAgent);
-      if (url.endsWith('/session')) return { data: { id: 'ses_test' } };
-      return { data: { parts: [{ type: 'text', text: 'ok' }] } };
-    });
-
-    await chat('sys', 'user');
-
-    assert.equal(agents.length, 2);
-    assert.equal(agents[0].options.rejectUnauthorized, false);
-    assert.equal(agents[1].options.rejectUnauthorized, false);
-  });
-
-  it('extracts text from OpenCode message parts', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    let calls = 0;
-    mock.method(axios, 'post', async () => {
-      calls += 1;
-      if (calls === 1) return { data: { id: 'ses_test' } };
-      return { data: { parts: [{ type: 'text', text: 'hello' }, { type: 'text', text: ' world' }] } };
-    });
+  it('can force opencode with AI_ASSISTANT_CLI', async () => {
+    const { argsPath } = await installFakeCLI('opencode');
+    process.env.AI_ASSISTANT_CLI = 'opencode';
+    process.env.MODEL = 'google/gemini-2.5-pro';
+    process.env.FAKE_AI_STDOUT = 'ok';
+    process.env.FAKE_AI_ARGS_PATH = argsPath;
 
     const result = await chat('sys', 'user');
 
-    assert.equal(result, 'hello world');
+    assert.equal(result, 'ok');
+    assert.match(await readFile(argsPath, 'utf8'), /run --model google\/gemini-2.5-pro/);
   });
 
-  it('throws an error when OpenCode fails instead of exiting the process', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_RETRY_ATTEMPTS = '1';
-    mock.method(axios, 'post', async () => {
-      const err = new Error('Request failed with status code 500');
-      err.response = { status: 500, data: { error: 'provider overloaded' } };
-      throw err;
-    });
+  it('throws an error when the CLI fails instead of exiting the process', async () => {
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_EXIT = '2';
+    process.env.FAKE_AI_STDERR = 'provider overloaded';
     const exitMock = mock.method(process, 'exit', () => { throw new Error('exit should not be called'); });
 
-    await assert.rejects(() => chat('sys', 'user'), /HTTP 500.*provider overloaded/);
+    await assert.rejects(() => chat('sys', 'user'), /provider overloaded/);
 
     assert.equal(exitMock.mock.calls.length, 0);
   });
-
-  it('retries transient OpenCode failures before returning content', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_RETRY_ATTEMPTS = '2';
-    let messageAttempts = 0;
-    mock.method(axios, 'post', async (url) => {
-      if (url.endsWith('/session')) return { data: { id: 'ses_test' } };
-      messageAttempts += 1;
-      if (messageAttempts === 1) {
-        const err = new Error('Request failed with status code 500');
-        err.response = { status: 500, data: { error: 'temporary failure' } };
-        throw err;
-      }
-      return { data: { parts: [{ type: 'text', text: 'ok after retry' }] } };
-    });
-
-    const result = await chat('sys', 'user');
-
-    assert.equal(result, 'ok after retry');
-    assert.equal(messageAttempts, 2);
-  });
 });
 
 describe('chatJSON', async () => {
   const { chatJSON } = await import('../llm.js');
 
   it('parses plain JSON response', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mockOpenCodeResponse('[{"level":"critical"}]');
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_STDOUT = '[{"level":"critical"}]';
 
     const result = await chatJSON('sys', 'user');
 
@@ -135,8 +98,8 @@ describe('chatJSON', async () => {
   });
 
   it('strips markdown code block before parsing', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mockOpenCodeResponse('```json\n[{"level":"info"}]\n```');
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_STDOUT = '```json\n[{"level":"info"}]\n```';
 
     const result = await chatJSON('sys', 'user');
 
@@ -144,8 +107,8 @@ describe('chatJSON', async () => {
   });
 
   it('extracts JSON array from surrounding prose', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mockOpenCodeResponse('**Reviewing findings**\n\n[{"level":"warning","suggestion":"x"}]\n\nDone.');
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_STDOUT = '**Reviewing findings**\n\n[{"level":"warning","suggestion":"x"}]\n\nDone.';
 
     const result = await chatJSON('sys', 'user');
 
@@ -153,8 +116,8 @@ describe('chatJSON', async () => {
   });
 
   it('extracts JSON object from surrounding prose', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mockOpenCodeResponse('**Begin Combine**\n{"merged_text":"repo block\\n\\nsource block"}');
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_STDOUT = '**Begin Combine**\n{"merged_text":"repo block\\n\\nsource block"}';
 
     const result = await chatJSON('sys', 'user');
 
@@ -162,8 +125,8 @@ describe('chatJSON', async () => {
   });
 
   it('returns [] when JSON is invalid', async () => {
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    mockOpenCodeResponse('not json');
+    await installFakeCLI('codex');
+    process.env.FAKE_AI_STDOUT = 'not json';
 
     const result = await chatJSON('sys', 'user');
 
diff --git a/app/test/preflight.test.js b/app/test/preflight.test.js
index 8d02e5e..4f6e4c6 100644
--- a/app/test/preflight.test.js
+++ b/app/test/preflight.test.js
@@ -1,21 +1,38 @@
 import { describe, it, afterEach, mock } from 'node:test';
 import assert from 'node:assert/strict';
 import axios from 'axios';
+import { mkdtemp, writeFile, chmod, rm } from 'fs/promises';
+import { tmpdir } from 'os';
+import { join } from 'path';
 import { checkRequiredEnv, verifyGiteaToken, verifyCommentToken, verifyLLM, runPreflight } from '../preflight.js';
 
 const LLM_ENV_KEYS = [
-  'OPENCODE_BASE_URL', 'OPENCODE_MODEL', 'OPENCODE_PROVIDER',
+  'AI_ASSISTANT_CLI', 'MODEL', 'OPENCODE_MODEL', 'PATH',
 ];
+const ORIGINAL_PATH = process.env.PATH;
 
 function clearLLMEnv() {
   for (const k of LLM_ENV_KEYS) delete process.env[k];
 }
 
-afterEach(() => {
+let tempDir;
+
+afterEach(async () => {
   mock.restoreAll();
   clearLLMEnv();
+  process.env.PATH = ORIGINAL_PATH;
+  if (tempDir) await rm(tempDir, { recursive: true, force: true });
+  tempDir = null;
 });
 
+async function installFakeCLI(command = 'codex') {
+  tempDir = await mkdtemp(join(tmpdir(), 'preflight-cli-test-'));
+  const script = join(tempDir, command);
+  await writeFile(script, '#!/bin/sh\nexit 0\n');
+  await chmod(script, 0o755);
+  process.env.PATH = tempDir;
+}
+
 describe('checkRequiredEnv', () => {
   it('reports all three missing when nothing provided', () => {
     const result = checkRequiredEnv({ token: '', repo: '', pr: '' });
@@ -102,81 +119,40 @@ describe('verifyCommentToken', () => {
 });
 
 describe('verifyLLM', () => {
-  it('fails when OpenCode is not configured', async () => {
+  it('fails when no supported assistant CLI is detected', async () => {
     clearLLMEnv();
+    process.env.AI_ASSISTANT_CLI = 'no-such-ai-cli';
+    process.env.PATH = '';
 
     const result = await verifyLLM();
 
     assert.equal(result.ok, false);
-    assert.match(result.error, /OPENCODE_BASE_URL/);
+    assert.match(result.error, /AI 助理 CLI/);
   });
 
-  it('checks OpenCode server provider and model', async () => {
+  it('passes when a supported assistant CLI is detected', async () => {
     clearLLMEnv();
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_PROVIDER = 'google';
-    process.env.OPENCODE_MODEL = 'gemini-2.5-flash';
-    const urls = [];
-    mock.method(axios, 'get', async (url) => {
-      urls.push(url);
-      if (url.endsWith('/global/health')) return { data: { healthy: true, version: '1.17.7' } };
-      return { data: { providers: [{ id: 'google', models: { 'gemini-2.5-flash': { id: 'gemini-2.5-flash' } } }] } };
-    });
+    await installFakeCLI('codex');
+    process.env.AI_ASSISTANT_CLI = 'codex';
+    process.env.MODEL = 'gpt-5-mini';
 
     const result = await verifyLLM();
 
     assert.equal(result.ok, true);
-    assert.equal(result.provider, 'opencode');
-    assert.deepEqual(urls, ['http://opencode.local:4096/global/health', 'http://opencode.local:4096/config/providers']);
+    assert.equal(result.provider, 'codex');
+    assert.equal(result.command, 'codex');
+    assert.equal(result.model, 'gpt-5-mini');
   });
 
-  it('fails when configured provider is missing', async () => {
+  it('fails when a requested CLI is not installed', async () => {
     clearLLMEnv();
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_PROVIDER = 'google';
-    mock.method(axios, 'get', async (url) => {
-      if (url.endsWith('/global/health')) return { data: { healthy: true } };
-      return { data: { providers: [{ id: 'anthropic', models: {} }] } };
-    });
+    process.env.AI_ASSISTANT_CLI = 'missing-cli';
+    process.env.PATH = '';
 
     const result = await verifyLLM();
 
     assert.equal(result.ok, false);
-    assert.match(result.error, /未設定 provider=google/);
-  });
-
-  it('fails when configured model is missing', async () => {
-    clearLLMEnv();
-    process.env.OPENCODE_BASE_URL = 'http://opencode.local:4096';
-    process.env.OPENCODE_PROVIDER = 'google';
-    process.env.OPENCODE_MODEL = 'gemini-2.5-pro';
-    mock.method(axios, 'get', async (url) => {
-      if (url.endsWith('/global/health')) return { data: { healthy: true } };
-      return { data: { providers: [{ id: 'google', models: { 'gemini-2.5-flash': { id: 'gemini-2.5-flash' } } }] } };
-    });
-
-    const result = await verifyLLM();
-
-    assert.equal(result.ok, false);
-    assert.match(result.error, /未列出 model=gemini-2.5-pro/);
-  });
-
-  it('passes an insecure https agent by default', async () => {
-    clearLLMEnv();
-    process.env.OPENCODE_BASE_URL = 'https://opencode.local:4096';
-    const agents = [];
-    mock.method(axios, 'get', async (url, opts) => {
-      agents.push(opts.httpsAgent);
-      if (url.endsWith('/global/health')) return { data: { healthy: true } };
-      return { data: { providers: [{ id: 'google', models: { 'gemini-2.5-flash': { id: 'gemini-2.5-flash' } } }] } };
-    });
-
-    const result = await verifyLLM();
-
-    assert.equal(result.ok, true);
-    assert.equal(agents.length, 2);
-    assert.equal(agents[0].options.rejectUnauthorized, false);
-    assert.equal(agents[1].options.rejectUnauthorized, false);
+    assert.match(result.error, /AI 助理 CLI/);
   });
 
 });
@@ -188,7 +164,7 @@ describe('runPreflight', () => {
       verifyToken: async () => ({ ok: true }),
       verifyComment: async () => ({ ok: true }),
       verifyRemote: () => ({ ok: true }),
-      verifyLLMFn: async () => ({ ok: true, provider: 'opencode' }),
+      verifyLLMFn: async () => ({ ok: true, provider: 'codex' }),
       ...overrides,
     };
   }
@@ -239,7 +215,7 @@ describe('runPreflight', () => {
 
   it('returns false when LLM verification fails', async () => {
     const result = await runPreflight('/ws', makeDeps({
-      verifyLLMFn: async () => ({ ok: false, error: 'OpenCode server 驗證失敗' }),
+      verifyLLMFn: async () => ({ ok: false, error: 'AI 助理 CLI 驗證失敗' }),
     }));
     assert.equal(result, false);
   });
-- 
2.53.0


From e17c25ce390e83df2af4615b450f45d36cce2e2d Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Sat, 27 Jun 2026 13:30:50 +0000
Subject: [PATCH 09/19] =?UTF-8?q?chore(ci):=20=E8=AA=BF=E6=95=B4=20AI=20?=
 =?UTF-8?q?=E5=AF=A9=E6=9F=A5=E5=B7=A5=E4=BD=9C=E6=B5=81=E7=A8=8B=E8=88=87?=
 =?UTF-8?q?=E6=A0=B9=E7=9B=AE=E9=8C=84=E6=B8=AC=E8=A9=A6=E6=8C=87=E4=BB=A4?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .gitea/workflows/ci.yaml | 21 ++++++++++++++++-----
 package.json             | 15 +++++++++++++++
 2 files changed, 31 insertions(+), 5 deletions(-)
 create mode 100644 package.json

diff --git a/.gitea/workflows/ci.yaml b/.gitea/workflows/ci.yaml
index 0377269..fd2171e 100644
--- a/.gitea/workflows/ci.yaml
+++ b/.gitea/workflows/ci.yaml
@@ -16,10 +16,22 @@ on:
     types: [opened, synchronize]
 # 定義此 workflow 的 jobs
 jobs:
+  release-tag-version:
+    name: Release Tag Version
+    runs-on: ubuntu
+    outputs:
+      version: ${{ steps.release-tag-version.outputs.version }}
+    steps:
+    - name: 計算版本號
+      id: release-tag-version
+      uses: https://gitea.jsc.idv.tw/composite-actions/release-tag-version@${{ vars.ACTION_RELEASE_TAG_VERSION }}
+      with:
+        is_beta: 'true'
   # job 識別碼：ai-code-review
   ai-code-review:
     # job 顯示名稱
     name: AI Code Review
+    needs: release-tag-version
     # 指定執行環境的 runner 標籤：ubuntu
     runs-on: ubuntu
     # 此 job 所需的權限設定
@@ -34,15 +46,14 @@ jobs:
     steps:
     - name: 取得程式碼
       uses: actions/checkout@${{ vars.ACTION_CHECKOUT_VERSION }}
-      # 步驟名稱：呼叫 OpenCode 進行 AI 程式碼審查
+    - name: 安裝 AI 工具
+      uses: https://gitea.jsc.idv.tw/docker-actions/antigravity@${{ vars.ACTION_ANTIGRAVITY_VERSION }}
     - name: AI 程式碼審查
       # 使用外部 composite action 執行審查邏輯
       # 版本由 repository variable ACTION_OPENCODE_CODE_REVIEW_VERSION 決定，便於集中管理版本
-      uses: ./
+      uses: https://gitea.jsc.idv.tw/docker-actions/ai-code-review@v${{ needs.release-tag-version.outputs.version }}
       # 傳遞給 composite action 的輸入參數
       with:
         # 留言用 token：取自 secret COMMENT_TOKEN，供 action 在 PR 上發布審查留言
         token: ${{ secrets.TOKEN }}
-        opencode_base_url: ${{ vars.OPENCODE_BASE_URL }}
-        opencode_provider: ${{ vars.OPENCODE_PROVIDER }}
-        opencode_model: ${{ vars.OPENCODE_PROVIDER_MODEL }}
+        model: ${{ vars.AI_CODE_REVIEW_MODEL }}
diff --git a/package.json b/package.json
new file mode 100644
index 0000000..2272d1b
--- /dev/null
+++ b/package.json
@@ -0,0 +1,15 @@
+{
+  "name": "ai-code-review",
+  "version": "1.0.0",
+  "description": "AI code review Gitea action",
+  "private": true,
+  "scripts": {
+    "test": "npm --prefix app test"
+  },
+  "repository": {
+    "type": "git",
+    "url": "https://gitea.jsc.idv.tw/docker-actions/ai-code-review.git"
+  },
+  "author": "Jeffery",
+  "license": "ISC"
+}
-- 
2.53.0


From df34376d9558941d87ac5a5c17378936d1886c76 Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Sat, 27 Jun 2026 13:38:31 +0000
Subject: [PATCH 10/19] =?UTF-8?q?chore(ci):=20=E5=A2=9E=E5=8A=A0=20AI=20?=
 =?UTF-8?q?=E5=8A=A9=E7=90=86=20CLI=20=E5=B7=A5=E5=85=B7=E6=AA=A2=E6=9F=A5?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .gitea/workflows/ci.yaml | 20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)

diff --git a/.gitea/workflows/ci.yaml b/.gitea/workflows/ci.yaml
index fd2171e..8f63b1f 100644
--- a/.gitea/workflows/ci.yaml
+++ b/.gitea/workflows/ci.yaml
@@ -48,9 +48,25 @@ jobs:
       uses: actions/checkout@${{ vars.ACTION_CHECKOUT_VERSION }}
     - name: 安裝 AI 工具
       uses: https://gitea.jsc.idv.tw/docker-actions/antigravity@${{ vars.ACTION_ANTIGRAVITY_VERSION }}
+    - name: 檢查 AI 助理 CLI
+      shell: bash
+      run: |
+        set -e
+        found=""
+        for tool in codex claude opencode; do
+          if command -v "$tool" >/dev/null 2>&1; then
+            found="${found} ${tool}"
+            "$tool" --version || true
+          fi
+        done
+        if [ -z "$found" ]; then
+          echo "找不到可用的 AI 助理 CLI（需要 codex、claude 或 opencode 其中之一）" >&2
+          exit 1
+        fi
+        echo "可用 AI 助理 CLI:${found}"
     - name: AI 程式碼審查
-      # 使用外部 composite action 執行審查邏輯
-      # 版本由 repository variable ACTION_OPENCODE_CODE_REVIEW_VERSION 決定，便於集中管理版本
+      # 使用外部 composite action 執行審查邏輯。
+      # 版本由 release-tag-version 計算，便於在 PR 中測試目前 action 版本。
       uses: https://gitea.jsc.idv.tw/docker-actions/ai-code-review@v${{ needs.release-tag-version.outputs.version }}
       # 傳遞給 composite action 的輸入參數
       with:
-- 
2.53.0


From 4818e82d765fc94913688316b7952a4cca24c4dd Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Sat, 27 Jun 2026 13:39:04 +0000
Subject: [PATCH 11/19] =?UTF-8?q?chore(ci):=20=E6=94=AF=E6=8F=B4=E8=A8=AD?=
 =?UTF-8?q?=E5=AE=9A=20AI=20=E5=8A=A9=E7=90=86=20CLI=20=E5=80=99=E9=81=B8?=
 =?UTF-8?q?=E5=B7=A5=E5=85=B7?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .gitea/workflows/ci.yaml | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/.gitea/workflows/ci.yaml b/.gitea/workflows/ci.yaml
index 8f63b1f..0d925bb 100644
--- a/.gitea/workflows/ci.yaml
+++ b/.gitea/workflows/ci.yaml
@@ -50,17 +50,20 @@ jobs:
       uses: https://gitea.jsc.idv.tw/docker-actions/antigravity@${{ vars.ACTION_ANTIGRAVITY_VERSION }}
     - name: 檢查 AI 助理 CLI
       shell: bash
+      env:
+        AI_ASSISTANT_CLI_TOOLS: ${{ vars.AI_ASSISTANT_CLI_TOOLS }}
       run: |
         set -e
+        tools="${AI_ASSISTANT_CLI_TOOLS:-codex claude opencode}"
         found=""
-        for tool in codex claude opencode; do
+        for tool in $tools; do
           if command -v "$tool" >/dev/null 2>&1; then
             found="${found} ${tool}"
             "$tool" --version || true
           fi
         done
         if [ -z "$found" ]; then
-          echo "找不到可用的 AI 助理 CLI（需要 codex、claude 或 opencode 其中之一）" >&2
+          echo "找不到可用的 AI 助理 CLI（候選工具：${tools}）" >&2
           exit 1
         fi
         echo "可用 AI 助理 CLI:${found}"
-- 
2.53.0


From 5457e680652c73be8b475947e5c471b58423ac56 Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Sat, 27 Jun 2026 13:41:15 +0000
Subject: [PATCH 12/19] =?UTF-8?q?chore(ci):=20=E6=94=B9=E5=9B=9E=E8=87=AA?=
 =?UTF-8?q?=E5=8B=95=E5=81=B5=E6=B8=AC=E5=B7=B2=E5=AE=89=E8=A3=9D=20AI=20?=
 =?UTF-8?q?=E5=8A=A9=E7=90=86=E5=B7=A5=E5=85=B7?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .gitea/workflows/ci.yaml | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/.gitea/workflows/ci.yaml b/.gitea/workflows/ci.yaml
index 0d925bb..8f63b1f 100644
--- a/.gitea/workflows/ci.yaml
+++ b/.gitea/workflows/ci.yaml
@@ -50,20 +50,17 @@ jobs:
       uses: https://gitea.jsc.idv.tw/docker-actions/antigravity@${{ vars.ACTION_ANTIGRAVITY_VERSION }}
     - name: 檢查 AI 助理 CLI
       shell: bash
-      env:
-        AI_ASSISTANT_CLI_TOOLS: ${{ vars.AI_ASSISTANT_CLI_TOOLS }}
       run: |
         set -e
-        tools="${AI_ASSISTANT_CLI_TOOLS:-codex claude opencode}"
         found=""
-        for tool in $tools; do
+        for tool in codex claude opencode; do
           if command -v "$tool" >/dev/null 2>&1; then
             found="${found} ${tool}"
             "$tool" --version || true
           fi
         done
         if [ -z "$found" ]; then
-          echo "找不到可用的 AI 助理 CLI（候選工具：${tools}）" >&2
+          echo "找不到可用的 AI 助理 CLI（需要 codex、claude 或 opencode 其中之一）" >&2
           exit 1
         fi
         echo "可用 AI 助理 CLI:${found}"
-- 
2.53.0


From 6f997083efd02c8366131334cde952d5e8690aa3 Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Sat, 27 Jun 2026 13:47:44 +0000
Subject: [PATCH 13/19] =?UTF-8?q?feat(ai-code-review):=20=E6=94=AF?=
 =?UTF-8?q?=E6=8F=B4=20Antigravity=20CLI?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .gitea/workflows/ci.yaml |  5 +++--
 app/config.js            | 18 ++++++++++++++++--
 app/llm.js               | 11 +++++++----
 app/preflight.js         |  2 +-
 app/test/config.test.js  | 13 ++++++++++++-
 app/test/llm.test.js     | 21 +++++++++++++++++++--
 app/test/usage.test.js   |  1 +
 app/usage.js             |  1 +
 8 files changed, 60 insertions(+), 12 deletions(-)

diff --git a/.gitea/workflows/ci.yaml b/.gitea/workflows/ci.yaml
index 8f63b1f..c5155fc 100644
--- a/.gitea/workflows/ci.yaml
+++ b/.gitea/workflows/ci.yaml
@@ -52,15 +52,16 @@ jobs:
       shell: bash
       run: |
         set -e
+        tools="$(node --input-type=module -e "import { getLLMCLICommands } from './app/config.js'; console.log(getLLMCLICommands().join(' '));")"
         found=""
-        for tool in codex claude opencode; do
+        for tool in $tools; do
           if command -v "$tool" >/dev/null 2>&1; then
             found="${found} ${tool}"
             "$tool" --version || true
           fi
         done
         if [ -z "$found" ]; then
-          echo "找不到可用的 AI 助理 CLI（需要 codex、claude 或 opencode 其中之一）" >&2
+          echo "找不到可用的 AI 助理 CLI（需要 ${tools} 其中之一）" >&2
           exit 1
         fi
         echo "可用 AI 助理 CLI:${found}"
diff --git a/app/config.js b/app/config.js
index e76389c..5efc5b4 100644
--- a/app/config.js
+++ b/app/config.js
@@ -42,6 +42,16 @@ const CLI_CANDIDATES = [
     command: 'claude',
     defaultModel: 'sonnet',
   },
+  {
+    provider: 'antigravity',
+    command: 'agy',
+    defaultModel: 'gemini-2.5-flash',
+  },
+  {
+    provider: 'antigravity',
+    command: 'antigravity',
+    defaultModel: 'gemini-2.5-flash',
+  },
   {
     provider: 'opencode',
     command: 'opencode',
@@ -49,6 +59,10 @@ const CLI_CANDIDATES = [
   },
 ];
 
+export function getLLMCLICommands() {
+  return CLI_CANDIDATES.map(c => c.command);
+}
+
 function commandExists(command) {
   try {
     execFileSync('/bin/sh', ['-lc', `command -v ${command}`], { stdio: 'ignore' });
@@ -61,11 +75,11 @@ function commandExists(command) {
 /**
  * 依環境變數解析並回傳 LLM 提供者設定。
  *
- * 優先使用 `AI_ASSISTANT_CLI` 指定的 CLI；未指定時依序偵測 codex、claude、opencode。
+ * 優先使用 `AI_ASSISTANT_CLI` 指定的 CLI；未指定時依序偵測 codex、claude、antigravity、opencode。
  * model 優先取 `MODEL`，再相容舊的 `OPENCODE_MODEL`，最後使用各 CLI 預設值。
  *
  * @param {{ commandExistsFn?: (command: string) => boolean }} [deps] - 可注入的 CLI 偵測函式，供測試使用。
- * @returns {{ provider: ('codex'|'claude'|'opencode'|null), apiKeys: string[], baseURL: null, model: (string|null), command: (string|null) }}
+ * @returns {{ provider: ('codex'|'claude'|'antigravity'|'opencode'|null), apiKeys: string[], baseURL: null, model: (string|null), command: (string|null) }}
  *          LLM 設定物件；`provider` 為 `null` 表示沒有可用的提供者。
  */
 export function getLLMConfig({ commandExistsFn = commandExists } = {}) {
diff --git a/app/llm.js b/app/llm.js
index 7a27866..470439f 100644
--- a/app/llm.js
+++ b/app/llm.js
@@ -23,13 +23,16 @@ function buildPrompt(systemPrompt, userContent) {
   ].join('\n');
 }
 
-function cliArgs(provider, model, promptFile = null) {
+function cliArgs(provider, model, promptFile = null, prompt = null) {
   if (provider === 'codex') {
     return ['exec', '--model', model, '--sandbox', 'read-only', '--ask-for-approval', 'never', '--skip-git-repo-check', '-'];
   }
   if (provider === 'claude') {
     return ['--print', '--model', model, '--permission-mode', 'dontAsk', '--no-session-persistence'];
   }
+  if (provider === 'antigravity') {
+    return ['-p', prompt, '--model', model];
+  }
   if (provider === 'opencode') {
     return ['run', '--model', model, '--format', 'default', '--file', promptFile, '請依附件 prompt.md 的完整內容執行，並只輸出要求的最終結果。'];
   }
@@ -50,7 +53,7 @@ async function runAssistantCLI({ provider, command, model }, prompt) {
     promptFile = join(tempDir, 'prompt.md');
     await writeFile(promptFile, prompt);
   }
-  const args = cliArgs(provider, model, promptFile);
+  const args = cliArgs(provider, model, promptFile, prompt);
   const maxBuffer = Number(process.env.AI_ASSISTANT_MAX_BUFFER || 20 * 1024 * 1024);
   const timeout = Number(process.env.AI_ASSISTANT_TIMEOUT_MS || 15 * 60 * 1000);
   try {
@@ -86,7 +89,7 @@ async function runAssistantCLI({ provider, command, model }, prompt) {
         if (code === 0) resolve(stdout.trim());
         else reject(Object.assign(new Error(`${provider} CLI exited with ${code ?? signal}`), { stdout, stderr }));
       });
-      child.stdin.end(provider === 'opencode' ? '' : prompt);
+      child.stdin.end(provider === 'opencode' || provider === 'antigravity' ? '' : prompt);
     });
   } finally {
     if (tempDir) await rm(tempDir, { recursive: true, force: true });
@@ -107,7 +110,7 @@ async function runAssistantCLI({ provider, command, model }, prompt) {
 export async function chat(systemPrompt, userContent) {
   const cfg = getLLMConfig();
   const { provider, command, model } = cfg;
-  if (!provider || !command) throw new Error('未偵測到可用 AI 助理 CLI，請安裝 codex、claude 或 opencode');
+  if (!provider || !command) throw new Error('未偵測到可用 AI 助理 CLI，請安裝 codex、claude、antigravity 或 opencode');
 
   line(`[LLM] provider=${provider} command=${command} model=${model}`);
 
diff --git a/app/preflight.js b/app/preflight.js
index ca7586d..e707238 100644
--- a/app/preflight.js
+++ b/app/preflight.js
@@ -106,7 +106,7 @@ export async function verifyCommentToken(token = GITEA_COMMENT_TOKEN) {
  */
 export async function verifyLLM() {
   const { provider, command, model } = getLLMConfig();
-  if (!provider || !command) return { ok: false, error: '未偵測到可用 AI 助理 CLI，請安裝 codex、claude 或 opencode' };
+  if (!provider || !command) return { ok: false, error: '未偵測到可用 AI 助理 CLI，請安裝 codex、claude、antigravity 或 opencode' };
   if (!model) return { ok: false, provider, error: '未設定 MODEL' };
   return { ok: true, provider, command, model };
 }
diff --git a/app/test/config.test.js b/app/test/config.test.js
index 942acc3..986b349 100644
--- a/app/test/config.test.js
+++ b/app/test/config.test.js
@@ -1,6 +1,6 @@
 import { describe, it, beforeEach, afterEach } from 'node:test';
 import assert from 'node:assert/strict';
-import { getLLMConfig, getOpenCodeHttpsAgent } from '../config.js';
+import { getLLMCLICommands, getLLMConfig, getOpenCodeHttpsAgent } from '../config.js';
 
 const ENV_KEYS = [
   'AI_ASSISTANT_CLI', 'MODEL', 'OPENCODE_MODEL',
@@ -19,6 +19,10 @@ afterEach(() => {
 });
 
 describe('getLLMConfig', () => {
+  it('exports the supported assistant CLI commands', () => {
+    assert.deepEqual(getLLMCLICommands(), ['codex', 'claude', 'agy', 'antigravity', 'opencode']);
+  });
+
   it('returns null provider when no env vars set', () => {
     const cfg = getLLMConfig({ commandExistsFn: () => false });
     assert.equal(cfg.provider, null);
@@ -43,6 +47,13 @@ describe('getLLMConfig', () => {
     assert.equal(cfg.model, 'gpt-5-mini');
   });
 
+  it('detects Antigravity through the agy command', () => {
+    const cfg = getLLMConfig({ commandExistsFn: command => command === 'agy' });
+    assert.equal(cfg.provider, 'antigravity');
+    assert.equal(cfg.command, 'agy');
+    assert.equal(cfg.model, 'gemini-2.5-flash');
+  });
+
   it('can force a CLI with AI_ASSISTANT_CLI', () => {
     process.env.AI_ASSISTANT_CLI = 'opencode';
     process.env.OPENCODE_MODEL = 'google/gemini-2.5-pro';
diff --git a/app/test/llm.test.js b/app/test/llm.test.js
index 5d7bd5c..0864089 100644
--- a/app/test/llm.test.js
+++ b/app/test/llm.test.js
@@ -31,13 +31,13 @@ async function installFakeCLI(command = 'codex') {
   const script = join(tempDir, command);
   await writeFile(script, `#!/bin/sh
 if [ -n "$FAKE_AI_ARGS_PATH" ]; then printf '%s\\n' "$*" > "$FAKE_AI_ARGS_PATH"; fi
-if [ -n "$FAKE_AI_STDIN_PATH" ]; then cat > "$FAKE_AI_STDIN_PATH"; else cat >/dev/null; fi
+if [ -n "$FAKE_AI_STDIN_PATH" ]; then /bin/cat > "$FAKE_AI_STDIN_PATH"; else /bin/cat >/dev/null; fi
 if [ -n "$FAKE_AI_STDERR" ]; then printf '%s' "$FAKE_AI_STDERR" >&2; fi
 if [ -n "$FAKE_AI_STDOUT" ]; then printf '%s' "$FAKE_AI_STDOUT"; fi
 exit "\${FAKE_AI_EXIT:-0}"
 `);
   await chmod(script, 0o755);
-  process.env.PATH = `${tempDir}:${saved.PATH || ''}`;
+  process.env.PATH = tempDir;
   return { stdinPath: join(tempDir, 'stdin.txt'), argsPath: join(tempDir, 'args.txt') };
 }
 
@@ -73,6 +73,23 @@ describe('chat - assistant CLI', async () => {
     assert.match(await readFile(argsPath, 'utf8'), /run --model google\/gemini-2.5-pro/);
   });
 
+  it('runs Antigravity through agy with MODEL and prompt argument', async () => {
+    const { stdinPath, argsPath } = await installFakeCLI('agy');
+    process.env.AI_ASSISTANT_CLI = 'agy';
+    process.env.MODEL = 'gemini-2.5-pro';
+    process.env.FAKE_AI_STDOUT = 'antigravity response';
+    process.env.FAKE_AI_STDIN_PATH = stdinPath;
+    process.env.FAKE_AI_ARGS_PATH = argsPath;
+
+    const result = await chat('sys', 'user');
+
+    assert.equal(result, 'antigravity response');
+    const args = await readFile(argsPath, 'utf8');
+    assert.match(args, /-p .*--model gemini-2.5-pro/s);
+    assert.match(args, /<system>\nsys\n<\/system>/);
+    assert.equal(await readFile(stdinPath, 'utf8'), '');
+  });
+
   it('throws an error when the CLI fails instead of exiting the process', async () => {
     await installFakeCLI('codex');
     process.env.FAKE_AI_EXIT = '2';
diff --git a/app/test/usage.test.js b/app/test/usage.test.js
index d6b5f4e..6ce0c95 100644
--- a/app/test/usage.test.js
+++ b/app/test/usage.test.js
@@ -120,6 +120,7 @@ describe('fetchAccountQuota', () => {
   it('reports 不適用 for local platforms', async () => {
     assert.equal((await fetchAccountQuota('ollama', {})).available, false);
     assert.equal((await fetchAccountQuota('opencode', {})).available, false);
+    assert.equal((await fetchAccountQuota('antigravity', {})).available, false);
   });
 
   it('degrades gracefully when the quota call throws', async () => {
diff --git a/app/usage.js b/app/usage.js
index 97eb2df..3e37821 100644
--- a/app/usage.js
+++ b/app/usage.js
@@ -197,6 +197,7 @@ const QUOTA_STRATEGIES = {
     return { available: false, reason: 'OpenAI 帳號額度需 dashboard session 權限，API key 無法取得' };
   },
   claude: async () => ({ available: false, reason: 'Anthropic 額度需 Admin API 權限，一般 API key 無法取得' }),
+  antigravity: async () => ({ available: false, reason: 'Antigravity 額度由 Google 帳務/方案管理，CLI 無法直接查詢' }),
   gemini: async () => ({ available: false, reason: 'Gemini 額度由 Google Cloud quota 管理，API key 無法直接查詢' }),
   amazonq: async () => ({ available: false, reason: 'Amazon Q 額度由 AWS 帳務管理，需 AWS 憑證查詢' }),
   ollama: async () => ({ available: false, reason: '本地服務，無帳號額度概念' }),
-- 
2.53.0


From d0c1bb0c202616294d2d48e00d36944ed1a50889 Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Sat, 27 Jun 2026 15:35:03 +0000
Subject: [PATCH 14/19] =?UTF-8?q?chore(ci):=20=E6=94=B9=E7=94=A8=20composi?=
 =?UTF-8?q?te-actions=20=E7=9A=84=20antigravity=20action?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .gitea/workflows/ci.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.gitea/workflows/ci.yaml b/.gitea/workflows/ci.yaml
index c5155fc..229121e 100644
--- a/.gitea/workflows/ci.yaml
+++ b/.gitea/workflows/ci.yaml
@@ -47,7 +47,7 @@ jobs:
     - name: 取得程式碼
       uses: actions/checkout@${{ vars.ACTION_CHECKOUT_VERSION }}
     - name: 安裝 AI 工具
-      uses: https://gitea.jsc.idv.tw/docker-actions/antigravity@${{ vars.ACTION_ANTIGRAVITY_VERSION }}
+      uses: https://gitea.jsc.idv.tw/composite-actions/antigravity@${{ vars.ACTION_ANTIGRAVITY_VERSION }}
     - name: 檢查 AI 助理 CLI
       shell: bash
       run: |
-- 
2.53.0


From ea2ac46d6c73629f78a8191c7753c47c1e52630f Mon Sep 17 00:00:00 2001
From: Jeffery <jiantw83@yahoo.com>
Date: Sat, 27 Jun 2026 15:56:12 +0000
Subject: [PATCH 15/19] =?UTF-8?q?chore(ci):=20=E5=82=B3=E5=85=A5=20Antigra?=
 =?UTF-8?q?vity=20OAuth=20secret?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .gitea/workflows/ci.yaml | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/.gitea/workflows/ci.yaml b/.gitea/workflows/ci.yaml
index 229121e..11b61af 100644
--- a/.gitea/workflows/ci.yaml
+++ b/.gitea/workflows/ci.yaml
@@ -48,6 +48,8 @@ jobs:
       uses: actions/checkout@${{ vars.ACTION_CHECKOUT_VERSION }}
     - name: 安裝 AI 工具
       uses: https://gitea.jsc.idv.tw/composite-actions/antigravity@${{ vars.ACTION_ANTIGRAVITY_VERSION }}
+      with:
+        oauth: ${{ secrets.ANTIGRAVITY_OAUTH }}
     - name: 檢查 AI 助理 CLI
       shell: bash
       run: |
-- 
2.53.0


From 811f56d715db1b0d9889e0cc51858486019c218d Mon Sep 17 00:00:00 2001
From: Jeffery <h3285@evertrust.com.tw>
Date: Mon, 29 Jun 2026 15:24:05 +0800
Subject: [PATCH 16/19] =?UTF-8?q?chore(ci):=20=E6=94=B9=E7=94=A8=20codex?=
 =?UTF-8?q?=20composite=20action=20=E5=AE=89=E8=A3=9D=20AI=20=E5=B7=A5?=
 =?UTF-8?q?=E5=85=B7?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .gitea/workflows/ci.yaml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.gitea/workflows/ci.yaml b/.gitea/workflows/ci.yaml
index 11b61af..8b937d2 100644
--- a/.gitea/workflows/ci.yaml
+++ b/.gitea/workflows/ci.yaml
@@ -47,9 +47,9 @@ jobs:
     - name: 取得程式碼
       uses: actions/checkout@${{ vars.ACTION_CHECKOUT_VERSION }}
     - name: 安裝 AI 工具
-      uses: https://gitea.jsc.idv.tw/composite-actions/antigravity@${{ vars.ACTION_ANTIGRAVITY_VERSION }}
+      uses: https://gitea.jsc.idv.tw/composite-actions/codex@${{ vars.ACTION_CODEX_VERSION }}
       with:
-        oauth: ${{ secrets.ANTIGRAVITY_OAUTH }}
+        oauth: ${{ secrets.CODEX_OAUTH }}
     - name: 檢查 AI 助理 CLI
       shell: bash
       run: |
-- 
2.53.0


From 34a9eb6d4e414a37d763b35cf6cb126c8a74f1f6 Mon Sep 17 00:00:00 2001
From: Jeffery <h3285@evertrust.com.tw>
Date: Mon, 29 Jun 2026 16:01:35 +0800
Subject: [PATCH 17/19] =?UTF-8?q?fix(llm):=20=E7=A7=BB=E9=99=A4=20codex=20?=
 =?UTF-8?q?exec=20=E5=B7=B2=E4=B8=8D=E6=94=AF=E6=8F=B4=E7=9A=84=20--ask-fo?=
 =?UTF-8?q?r-approval=20=E5=8F=83=E6=95=B8?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 app/llm.js | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/app/llm.js b/app/llm.js
index 470439f..0226b0a 100644
--- a/app/llm.js
+++ b/app/llm.js
@@ -25,7 +25,7 @@ function buildPrompt(systemPrompt, userContent) {
 
 function cliArgs(provider, model, promptFile = null, prompt = null) {
   if (provider === 'codex') {
-    return ['exec', '--model', model, '--sandbox', 'read-only', '--ask-for-approval', 'never', '--skip-git-repo-check', '-'];
+    return ['exec', '--model', model, '--sandbox', 'read-only', '--skip-git-repo-check', '-'];
   }
   if (provider === 'claude') {
     return ['--print', '--model', model, '--permission-mode', 'dontAsk', '--no-session-persistence'];
-- 
2.53.0


From 9ec30e1abbf43029519dafc712b17cfdf008a48b Mon Sep 17 00:00:00 2001
From: Jeffery <h3285@evertrust.com.tw>
Date: Mon, 29 Jun 2026 16:01:38 +0800
Subject: [PATCH 18/19] =?UTF-8?q?chore(config):=20=E5=B0=87=20codex=20?=
 =?UTF-8?q?=E9=A0=90=E8=A8=AD=E6=A8=A1=E5=9E=8B=E6=9B=B4=E6=96=B0=E7=82=BA?=
 =?UTF-8?q?=20gpt-5.4-mini?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 app/config.js | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/app/config.js b/app/config.js
index 5efc5b4..4c8db14 100644
--- a/app/config.js
+++ b/app/config.js
@@ -35,7 +35,7 @@ const CLI_CANDIDATES = [
   {
     provider: 'codex',
     command: 'codex',
-    defaultModel: 'gpt-5',
+    defaultModel: 'gpt-5.4-mini',
   },
   {
     provider: 'claude',
-- 
2.53.0


From d0b22e8444061ff19c4355606554e7189d81286a Mon Sep 17 00:00:00 2001
From: AI Review Bot <ai-review[bot]@gitea>
Date: Mon, 29 Jun 2026 08:07:59 +0000
Subject: [PATCH 19/19] chore: update ai-review findings
 [ai-review-bot][success]

---
 .gitea/ai-review/findings.json | 67 +++++++++++++++++++++++++++++++++-
 1 file changed, 66 insertions(+), 1 deletion(-)

diff --git a/.gitea/ai-review/findings.json b/.gitea/ai-review/findings.json
index fe51488..92d2571 100644
--- a/.gitea/ai-review/findings.json
+++ b/.gitea/ai-review/findings.json
@@ -1 +1,66 @@
-[]
+[
+  {
+    "level": "warning",
+    "role": "Assassin",
+    "location": "app/git.js:67",
+    "problem": "這裡對所有 git 網路操作強制關閉 SSL 驗證。攻擊者若能攔截遠端回應，就能偽造 `ls-remote`、污染抓到的內容，甚至在有認證的情況下竊取 GITEA_TOKEN，讓後續審查建立在被污染的資料上。",
+    "suggestion": "移除 `GIT_SSL_NO_VERIFY=true`。若是內部憑證問題，改用正確的 CA bundle 或指定受信任憑證檔，不要整體關閉 git 的憑證驗證。",
+    "is_new": true
+  },
+  {
+    "level": "warning",
+    "role": "Assassin",
+    "location": "app/usage.js:180",
+    "problem": "這裡對 OpenRouter 額度查詢也明確套用不驗證憑證的 agent。攻擊者只要攔截這條 HTTPS 請求，就能讀到 `Authorization: Bearer` 的 API key，還能回傳假 quota 資料誤導流程判斷。",
+    "suggestion": "改回預設的 HTTPS 憑證驗證。若必須支援自簽憑證，僅對特定內部主機單獨配置受信任 CA，且不要對外部 API 一律關閉驗證。",
+    "is_new": true
+  },
+  {
+    "level": "warning",
+    "role": "Bard",
+    "location": "action.yaml:8",
+    "problem": "這個新加入的 `model` 輸入沒有任何描述，讓 action 的公開介面像是少了說明書的樂譜，讀者只能靠猜來理解它要吃什麼值。",
+    "suggestion": "補上一句簡短但明確的描述，例如說明這個欄位是「要使用的模型名稱」或「AI 助理 CLI 的模型識別字串」；讓使用者一眼看懂用途。",
+    "is_new": true
+  },
+  {
+    "level": "warning",
+    "role": "Bard",
+    "location": "app/config.js:4",
+    "problem": "`process.env.NODE_TLS_REJECT_UNAUTHORIZED = '0'` 放在模組載入時就直接生效，這種全域副作用太早出聲，會讓讀者在還沒看到任何呼叫前，就先被迫接受一個隱晦的環境改寫。",
+    "suggestion": "把這種全域設定移到明確的啟動流程或初始化函式中，至少搭配更醒目的註解，讓副作用出現的位置和目的都更清楚。",
+    "is_new": true
+  },
+  {
+    "level": "warning",
+    "role": "Bard",
+    "location": "app/config.js:25",
+    "problem": "`getInsecureHttpsAgent` 與 `getOpenCodeHttpsAgent` 同時存在，卻指向同一個物件來源，等於替同一段旋律寫了兩個名字，容易讓後續維護者搞不清楚哪個才是正規稱呼。",
+    "suggestion": "保留一個主名稱即可；如果一定要兼容舊名稱，請加上明確的遷移註解，說明哪個是新名字、哪個只是過渡別名。",
+    "is_new": true
+  },
+  {
+    "level": "warning",
+    "role": "Bard",
+    "location": "app/llm.js:27",
+    "problem": "`cliArgs(provider, model, promptFile = null, prompt = null)` 這個簽名把不同 provider 的參數意義混在一起，光看位置很難分辨哪個值在某條分支才有用，讀起來像把三種節拍硬塞進同一小節。",
+    "suggestion": "改成物件參數或拆成各 provider 專用的組裝函式，讓每個欄位的用途能被名字直接說明，而不是靠呼叫順序猜測。",
+    "is_new": true
+  },
+  {
+    "level": "warning",
+    "role": "Bard",
+    "location": "app/preflight.js:103",
+    "problem": "`verifyLLM` 的 JSDoc 仍然只描述 `provider` 與 `error`，但實際回傳已經多了 `command` 和 `model`。註解和程式碼不同步，會讓人以為這段函式比實際更單薄。",
+    "suggestion": "把 `@returns` 的型別與欄位補完整，讓文件精準反映目前回傳結構，避免讀者依舊沿用舊版心智模型。",
+    "is_new": true
+  },
+  {
+    "level": "info",
+    "role": "Bard",
+    "location": "app/preflight.js:170",
+    "problem": "成功訊息仍寫成 `LLM provider=... CLI 可用`，語氣還停在 provider 中心，但整個流程已改成 CLI 中心，術語有些半新半舊，節奏不夠一致。",
+    "suggestion": "把成功訊息統一成 CLI 觀點，例如直接報 `command=...` 或改成更中性的「LLM CLI 可用」，讓文字系統保持同一把調性。",
+    "is_new": true
+  }
+]
-- 
2.53.0