fix: Codex 自动压缩不触发 —— 接管时链接模型目录(绝对路径)进 config.toml (hot 1.5.83)

2026-08-16 12:45:32 +08:00
parent e9a261482c
commit 1b687c01e9
50 changed files with 10885 additions and 1161 deletions
@@ -1,20 +1,21 @@
-// voice_cdp.js —— Codex 语音输入注入：主进程侧。
-// 用 CDP（Chrome DevTools Protocol）连接带 --remote-debugging-port 的 Codex 桌面版，
+// voice_cdp.js —— ChatGPT/Codex 语音输入注入：主进程侧。
+// 用 CDP（Chrome DevTools Protocol）连接带 --remote-debugging-port 的 ChatGPT/Codex 桌面版，
 // 往其页面注入麦克风按钮脚本，录音经 binding 回传 → 调中转站 Whisper 转写 → 文字回填 composer。
 // 不改 Codex 源码、不写补丁文件。无第三方依赖：CDP over WebSocket 用 Node 内置 net 手写最小 WS 客户端
 // （Electron 34 主进程的 Node 无全局 WebSocket）；HTTP 用全局 fetch；multipart 用全局 FormData/Blob。
 //
 // 机制参考 CodexPlusPlus 的 bridge.rs：Runtime.evaluate / Page.addScriptToEvaluateOnNewDocument /
 // Runtime.addBinding + Runtime.bindingCalled。
-const { spawn, execFile } = require("child_process");
-const os = require("os");
-const path = require("path");
+const { execFile, spawn } = require("child_process");
+const fs = require("fs");
 const net = require("net");
 const crypto = require("crypto");
+const os = require("os");
+const path = require("path");

 function dlog(...a) { try { console.log("[voice]", ...a); } catch {} }
 const router = require("./router");
-const providerSync = require("./provider_sync");
+const codexDesktop = require("./codex_desktop");
 const { INJECT_SCRIPT } = require("./voice_inject");
 const { TOKEN_INJECT_SCRIPT } = require("./token_inject");
 const tokenUsage = require("./token_usage");
@@ -25,28 +26,39 @@ const CDP_HTTP_TIMEOUT = 3000; // GET /json 超时
 const CDP_WS_TIMEOUT = 5000; // WebSocket 握手超时
 const CDP_CMD_TIMEOUT = 5000; // 单条 CDP 命令超时
 const READY_POLL_INTERVAL = 500; // 轮询就绪间隔
-const READY_POLL_MAX_MS = 20000; // 轮询就绪最长等待
+const READY_POLL_MAX_MS = 180000; // 轮询就绪最长等待：Windows 冷启动可能要 2～3 分钟
+const PACKAGED_READY_POLL_MAX_MS = 180000; // Store/MSIX 版 AppID 启动后也要给足白屏加载时间
+const WINDOWS_LAUNCH_TIMEOUT_MS = 4500;
+const WINDOWS_APPID_ACTIVATE_TIMEOUT_MS = 10000;
 const RECORD_MAX_MS = 60000; // 单次录音上限（注入脚本侧也限）

 // —— 状态机 ——
-// unknown / checking / not_running / running_no_port / restarting / connecting /
-// injecting / connected / disconnected / error
+// unknown / checking / not_running / running_no_port / restarting / waiting_debug_port /
+// unsupported_packaged / connecting / injecting / connected / disconnected / error
 let state = {
  phase: "unknown",
  port: DEFAULT_PORT,
  codexPath: "",
+  codexPackaged: false,
+  codexLaunchType: "exe",
+  codexAppId: "",
+  launchWarning: "",
+  portDiagnostic: "",
  lastError: "",
+  discovery: null,
  target: null, // { id, url, title }
 };
 let session = null; // 当前 CdpSession
+let voiceScriptIdentifier = ""; // 用于断开时撤销后续页面的语音自动注入
 let networkSessions = []; // worker 等附加 target 的 Network 监听 session
 let networkAttachGeneration = 0;
 let onStatusCb = null;
+let onSessionDeleteCb = null;
 let tokenSyncTimer = null;

 function snapshot() {
-  const { phase, port, codexPath, lastError, model, target } = state;
-  return { phase, port, codexPath, lastError, model, target };
+  const { phase, port, codexPath, codexPackaged, codexLaunchType, codexAppId, launchWarning, portDiagnostic, lastError, model, target, discovery } = state;
+  return { phase, port, codexPath, codexPackaged, codexLaunchType, codexAppId, launchWarning, portDiagnostic, lastError, model, target, discovery };
 }
 function setPhase(phase, extra = {}) {
  state = { ...state, phase, ...extra };
@@ -59,7 +71,7 @@ function setPhase(phase, extra = {}) {

 async function syncTokenUsage() {
  if (!session) return;
-  const usage = tokenUsage.readDaily();
+  const usage = await tokenUsage.readDaily();
  if (!usage || !usage.available) return;
  try {
    await session.send("Runtime.evaluate", {
@@ -81,36 +93,290 @@ function stopTokenSync() {
 }

 // —— 探测 Codex 可执行路径 ——
-function codexExecutable() {
-  const home = os.homedir();
-  if (process.platform === "darwin") {
-    const cands = [
-      path.join(home, "Applications/ChatGPT.app/Contents/MacOS/ChatGPT"),
-      "/Applications/ChatGPT.app/Contents/MacOS/ChatGPT",
-      "/Applications/Codex.app/Contents/MacOS/Codex",
-    ];
-    for (const p of cands) {
-      try {
-        if (require("fs").existsSync(p)) return p;
-      } catch {}
-    }
-    return "";
-  }
+function scanCodexExecutable() {
+  const candidates = codexDesktop.desktopExecutableCandidates();
+  const candidate = candidates.find((item) => item && item.path && !item.packaged) ||
+    candidates.find((item) => item && item.path && item.packaged);
+  const exe = candidate && candidate.path ? candidate.path : "";
+  state.codexPath = exe;
+  state.codexPackaged = !!(candidate && candidate.packaged);
+  state.codexLaunchType = (candidate && candidate.launchType) || "exe";
+  state.codexAppId = (candidate && candidate.appId) || "";
+  state.discovery = codexDesktop.discoveryReport ? codexDesktop.discoveryReport() : null;
+  return exe;
+}
+
+function desktopNotFoundError() {
+  const summary = codexDesktop.discoverySummary ? codexDesktop.discoverySummary() : "";
+  const discovery = state.discovery || {};
+  const hasPackaged = Array.isArray(discovery.candidates) && discovery.candidates.some((item) => item && item.packaged);
  if (process.platform === "win32") {
-    const local = process.env.LOCALAPPDATA || "";
-    const pf = process.env["ProgramFiles"] || "C:\\Program Files";
-    const cands = [
-      local && path.join(local, "Programs\\ChatGPT\\ChatGPT.exe"),
-      path.join(pf, "ChatGPT\\ChatGPT.exe"),
-    ].filter(Boolean);
-    for (const p of cands) {
-      try {
-        if (require("fs").existsSync(p)) return p;
-      } catch {}
-    }
-    return "";
+    return "未找到可直接启动的 ChatGPT/Codex 桌面程序。" +
+      (summary ? "扫描结果：" + summary + "。" : "") +
+      (hasPackaged ? "检测到 WindowsApps/Store 受保护入口，但它不能直接附加调试端口。" : "") +
+      "请安装官网下载的桌面版后重试。";
  }
-  return "";
+  return "未找到 ChatGPT/Codex 桌面程序。" + (summary ? "扫描结果：" + summary + "。" : "");
+}
+
+function desktopLaunchError(exePath, err) {
+  const msg = String((err && err.message) || err || "");
+  if (process.platform === "win32" && /eperm/i.test(msg)) {
+    return "Windows 拒绝启动 ChatGPT/Codex 调试模式（EPERM）。路径：" + exePath +
+      "。如果这是 Microsoft Store/WindowsApps 版本，请改装官网下载的桌面版；如果是普通 exe，请用管理员权限启动工具箱后再试。";
+  }
+  return "启动 ChatGPT/Codex 失败：" + msg + (exePath ? "。路径：" + exePath : "");
+}
+
+function compactError(err, fallback = "未知错误") {
+  const parts = [
+    err && err.stderr,
+    err && err.stdout,
+    err && err.message,
+    err,
+  ].filter(Boolean).map((v) => String(v));
+  const raw = parts.find((v) => v.trim()) || fallback;
+  return raw
+    .replace(/Command failed:[\s\S]*?(?=(\r?\n|$))/i, "")
+    .replace(/\s+/g, " ")
+    .trim()
+    .slice(0, 220) || fallback;
+}
+
+function compactCommandLine(cmd) {
+  const text = String(cmd || "").replace(/\s+/g, " ").trim();
+  if (!text) return "";
+  const hasWindowsApps = /WindowsApps/i.test(text);
+  const hasDebugArg = /--remote-debugging-port/i.test(text);
+  const exe = (text.match(/(?:^|[ "])([A-Z]:\\[^"]*?(?:ChatGPT|Codex)[^"\\]*\.exe)/i) || [])[1] || "";
+  const pidPart = exe ? exe.replace(/^.*[\\/]/, "") : "";
+  const tags = [];
+  if (hasWindowsApps) tags.push("WindowsApps");
+  tags.push(hasDebugArg ? "含调试参数" : "无调试参数");
+  return [pidPart, tags.join("/")].filter(Boolean).join(" ");
+}
+
+function updateWindowsProcessDiagnostic(port) {
+  if (process.platform !== "win32" || !codexDesktop.windowsProcessDiagnostics) return null;
+  try {
+    const d = codexDesktop.windowsProcessDiagnostics(port);
+    if (!d || !Array.isArray(d.rows) || !d.rows.length) {
+      state.portDiagnostic = "未检测到 ChatGPT/Codex 进程";
+      return d;
+    }
+    state.portDiagnostic = d.hasDebugArg
+      ? "进程命令行已包含调试端口参数"
+      : "进程命令行未包含 --remote-debugging-port=" + port + "，调试端口不会打开";
+    const rowsSummary = d.rows.map((row) => {
+      const name = (row && row.Name) || "process";
+      const pid = (row && row.ProcessId) || "?";
+      const cmd = compactCommandLine((row && row.CommandLine) || "");
+      return `${name}#${pid}${cmd ? " " + cmd : ""}`;
+    }).join(" | ");
+    if (rowsSummary) state.portDiagnostic += "；" + rowsSummary;
+    state.portDiagnostic = state.portDiagnostic.slice(0, 260);
+    return d;
+  } catch (e) {
+    state.portDiagnostic = "读取进程命令行失败：" + compactError(e);
+    return null;
+  }
+}
+
+function debugPortFailureMessage(action, port) {
+  const diag = updateWindowsProcessDiagnostic(port);
+  if (state.codexPackaged) {
+    const reason = diag && diag.rows && diag.rows.length && !diag.hasDebugArg
+      ? "当前 Store/MSIX 启动参数没有进入 Codex 进程，无法打开调试端口。"
+      : "当前 Store/MSIX 启动方式未开放调试端口。";
+    return "已尝试通过 Windows AppID " + action + " ChatGPT/Codex，但未检测到调试端口。" +
+      reason + "建议改装官网下载桌面版。";
+  }
+  return action === "重启" ? "重启 ChatGPT 后未开放调试端口" : "Codex 启动后未开放调试端口，请重试";
+}
+
+function isPackagedDebugArgMissing(port) {
+  if (!state.codexPackaged || process.platform !== "win32") return false;
+  const d = updateWindowsProcessDiagnostic(port);
+  return !!(d && Array.isArray(d.rows) && d.rows.length && !d.hasDebugArg);
+}
+
+function setDebugPortFailure(action, port) {
+  const unsupported = isPackagedDebugArgMissing(port);
+  setPhase(unsupported ? "unsupported_packaged" : "error", {
+    lastError: debugPortFailureMessage(action, port) + (state.launchWarning ? "（" + state.launchWarning + "）" : ""),
+    portDiagnostic: state.portDiagnostic,
+  });
+}
+
+function windowsLaunchArgs(port) {
+  return [
+    `--remote-debugging-port=${port}`,
+    `--remote-allow-origins=http://127.0.0.1:${port}`,
+  ];
+}
+
+function quoteWindowsArgument(arg) {
+  const text = String(arg || "");
+  if (text && !/[ \t"]/.test(text)) return text;
+  let out = "\"";
+  let backslashes = 0;
+  for (const ch of text) {
+    if (ch === "\\") {
+      backslashes += 1;
+      continue;
+    }
+    if (ch === "\"") {
+      out += "\\".repeat(backslashes * 2 + 1) + "\"";
+      backslashes = 0;
+      continue;
+    }
+    out += "\\".repeat(backslashes) + ch;
+    backslashes = 0;
+  }
+  out += "\\".repeat(backslashes * 2) + "\"";
+  return out;
+}
+
+function windowsCommandLineArguments(args) {
+  return args.map(quoteWindowsArgument).join(" ");
+}
+
+function windowsActivateAppId(appId, port) {
+  return new Promise((resolve, reject) => {
+    if (process.platform !== "win32") {
+      reject(new Error("AppID activation only supports Windows"));
+      return;
+    }
+    const script = [
+      "$ErrorActionPreference='Stop'",
+      "[Console]::OutputEncoding=[System.Text.Encoding]::UTF8",
+      "$OutputEncoding=[System.Text.Encoding]::UTF8",
+      "$code = @'",
+      "using System;",
+      "using System.Runtime.InteropServices;",
+      "namespace TdCodexAppActivation {",
+      "[ComImport, Guid(\"45BA127D-10A8-46EA-8AB7-56EA9078943C\")]",
+      "public class ApplicationActivationManager {}",
+      "[ComImport, Guid(\"2e941141-7f97-4756-ba1d-9decde894a3d\"), InterfaceType(ComInterfaceType.InterfaceIsIUnknown)]",
+      "public interface IApplicationActivationManager {",
+      "  [PreserveSig]",
+      "  int ActivateApplication([MarshalAs(UnmanagedType.LPWStr)] string appUserModelId, [MarshalAs(UnmanagedType.LPWStr)] string arguments, int options, out uint processId);",
+      "  [PreserveSig]",
+      "  int ActivateForFile([MarshalAs(UnmanagedType.LPWStr)] string appUserModelId, IntPtr itemArray, [MarshalAs(UnmanagedType.LPWStr)] string verb, out uint processId);",
+      "  [PreserveSig]",
+      "  int ActivateForProtocol([MarshalAs(UnmanagedType.LPWStr)] string appUserModelId, IntPtr itemArray, out uint processId);",
+      "}",
+      "public static class Launcher {",
+      "  const int COINIT_APARTMENTTHREADED = 0x2;",
+      "  const int RPC_E_CHANGED_MODE = unchecked((int)0x80010106);",
+      "  [DllImport(\"ole32.dll\")] static extern int CoInitializeEx(IntPtr pvReserved, int dwCoInit);",
+      "  [DllImport(\"ole32.dll\")] static extern void CoUninitialize();",
+      "  public static uint Activate(string appUserModelId, string arguments) {",
+      "    int coinit = CoInitializeEx(IntPtr.Zero, COINIT_APARTMENTTHREADED);",
+      "    bool shouldUninitialize = coinit >= 0;",
+      "    if (coinit < 0 && coinit != RPC_E_CHANGED_MODE) Marshal.ThrowExceptionForHR(coinit);",
+      "    try {",
+      "      var manager = (IApplicationActivationManager)new ApplicationActivationManager();",
+      "      uint processId = 0;",
+      "      int hr = manager.ActivateApplication(appUserModelId, arguments, 0, out processId);",
+      "      if (hr != 0) throw new COMException(\"ActivateApplication failed\", hr);",
+      "      return processId;",
+      "    } finally {",
+      "      if (shouldUninitialize) CoUninitialize();",
+      "    }",
+      "  }",
+      "}",
+      "}",
+      "'@",
+      "Add-Type -TypeDefinition $code -ErrorAction Stop",
+      "$activatedProcessId=[TdCodexAppActivation.Launcher]::Activate($env:TD_CODEX_APPID, $env:TD_CODEX_LAUNCH_ARGS)",
+      "[pscustomobject]@{ok=$true;pid=$activatedProcessId;args=$env:TD_CODEX_LAUNCH_ARGS}|ConvertTo-Json -Compress",
+    ].join("\n");
+    const scriptPath = path.join(os.tmpdir(), "td-codex-appid-activate-" + process.pid + "-" + Date.now() + ".ps1");
+    try {
+      fs.writeFileSync(scriptPath, script, "utf8");
+    } catch (e) {
+      reject(e);
+      return;
+    }
+    execFile("powershell.exe", ["-NoProfile", "-Sta", "-ExecutionPolicy", "Bypass", "-File", scriptPath], {
+      env: {
+        ...process.env,
+        TD_CODEX_APPID: String(appId || ""),
+        TD_CODEX_LAUNCH_PORT: String(port),
+        TD_CODEX_LAUNCH_ARGS: windowsCommandLineArguments(windowsLaunchArgs(port)),
+      },
+      windowsHide: true,
+      timeout: WINDOWS_APPID_ACTIVATE_TIMEOUT_MS,
+    }, (err, stdout) => {
+      try { fs.rmSync(scriptPath, { force: true }); } catch {}
+      if (err) {
+        reject(err);
+        return;
+      }
+      let data = null;
+      try {
+        data = JSON.parse(String(stdout || "").trim());
+      } catch {}
+      resolve(data || { ok: true });
+    });
+  });
+}
+
+function windowsShellStart(exePath, port) {
+  return new Promise((resolve, reject) => {
+    if (process.platform !== "win32") {
+      reject(new Error("shell start only supports Windows"));
+      return;
+    }
+    const launchType = state.codexLaunchType || "exe";
+    const target = launchType === "appId"
+      ? "shell:AppsFolder\\" + String(state.codexAppId || exePath || "").trim()
+      : String(exePath || "").replace(/^\\\\\?\\/, "");
+    if (launchType === "appId") {
+      windowsActivateAppId(state.codexAppId || exePath, port).then((data) => {
+        const pid = data && data.pid ? `，PID ${data.pid}` : "";
+        state.launchWarning = "已通过 Windows AppID 激活 Store/MSIX 版 Codex" + pid;
+        resolve(true);
+      }).catch((activateErr) => {
+        execFile("explorer.exe", [target], { windowsHide: true, timeout: WINDOWS_LAUNCH_TIMEOUT_MS }, (err) => {
+          const first = "COM 激活提示：" + compactError(activateErr, "AppID 激活失败");
+          if (err) {
+            // explorer.exe often returns a failure code for shell:AppsFolder
+            // even after Windows has already handed off and launched the app.
+            const second = "explorer 提示：" + compactError(err, "explorer 启动失败");
+            state.launchWarning = (first + "；" + second).slice(0, 220);
+            resolve(false);
+            return;
+          }
+          state.launchWarning = (first + "；已通过 explorer 兜底拉起 Store/MSIX 版 Codex").slice(0, 220);
+          resolve(true);
+        });
+      });
+      return;
+    }
+    const script = [
+      "$target=$env:TD_CODEX_LAUNCH_TARGET",
+      "$args=@('--remote-debugging-port=' + $env:TD_CODEX_LAUNCH_PORT, '--remote-allow-origins=http://127.0.0.1:' + $env:TD_CODEX_LAUNCH_PORT)",
+      "Start-Process -FilePath $target -ArgumentList $args",
+    ].join(";");
+    execFile("powershell.exe", ["-NoProfile", "-ExecutionPolicy", "Bypass", "-Command", script], {
+      env: {
+        ...process.env,
+        TD_CODEX_LAUNCH_TARGET: target,
+        TD_CODEX_LAUNCH_PORT: String(port),
+      },
+      windowsHide: true,
+      timeout: WINDOWS_LAUNCH_TIMEOUT_MS,
+    }, (err) => {
+      if (err) {
+        reject(err);
+        return;
+      }
+      state.launchWarning = "";
+      resolve(true);
+    });
+  });
 }

 // —— HTTP：GET /json 取 targets ——
@@ -427,7 +693,22 @@ async function handleBinding(params) {
  } catch {
    return;
  }
-  if (!payload || payload.kind !== "transcribe") return;
+  if (!payload) return;
+  if (
+    payload.kind === "delete-session" ||
+    payload.kind === "delete-session-clipboard-prepare" ||
+    payload.kind === "delete-session-clipboard-read"
+  ) {
+    if (onSessionDeleteCb) {
+      try { await onSessionDeleteCb(payload); } catch (e) { dlog("永久删除会话请求失败:", e && e.message); }
+    }
+    return;
+  }
+  if (payload.kind === "debug-sidebar-dump") {
+    dlog("侧边栏诊断:", JSON.stringify(payload.info || {}));
+    return;
+  }
+  if (payload.kind !== "transcribe") return;
  const { id, audioB64, mime } = payload;
  dlog("binding 收到转写请求 id=" + id, "mime=" + mime, "audioB64长度=" + (audioB64?audioB64.length:0));
  setPhase("transcribing");
@@ -456,6 +737,11 @@ function connectedPhase() {

 // —— 注入：addBinding + addScriptToEvaluateOnNewDocument + 立即 evaluate + Network 监听 ——
 async function inject(sessionObj) {
+  // 清除 require 缓存，使 voice_inject.js 的修改能在重连时生效，无需重启整个应用
+  delete require.cache[require.resolve("./voice_inject")];
+  delete require.cache[require.resolve("./token_inject")];
+  const { INJECT_SCRIPT: FRESH_INJECT } = require("./voice_inject");
+  const { TOKEN_INJECT_SCRIPT: FRESH_TOKEN } = require("./token_inject");
  await sessionObj.send("Runtime.enable");
  try {
    await sessionObj.send("Runtime.removeBinding", { name: BINDING_NAME });
@@ -463,11 +749,12 @@ async function inject(sessionObj) {
  await sessionObj.send("Runtime.addBinding", { name: BINDING_NAME });
  await sessionObj.send("Page.enable");
  // 语音注入脚本
-  await sessionObj.send("Page.addScriptToEvaluateOnNewDocument", { source: INJECT_SCRIPT });
-  await sessionObj.send("Runtime.evaluate", { expression: INJECT_SCRIPT, awaitPromise: false });
+  const voiceScript = await sessionObj.send("Page.addScriptToEvaluateOnNewDocument", { source: FRESH_INJECT });
+  voiceScriptIdentifier = (voiceScript && voiceScript.identifier) || "";
+  await sessionObj.send("Runtime.evaluate", { expression: FRESH_INJECT, awaitPromise: false });
  // Token 用量 UI 注入脚本（只负责 UI + localStorage，不管网络抓取）
-  await sessionObj.send("Page.addScriptToEvaluateOnNewDocument", { source: TOKEN_INJECT_SCRIPT });
-  await sessionObj.send("Runtime.evaluate", { expression: TOKEN_INJECT_SCRIPT, awaitPromise: false });
+  await sessionObj.send("Page.addScriptToEvaluateOnNewDocument", { source: FRESH_TOKEN });
+  await sessionObj.send("Runtime.evaluate", { expression: FRESH_TOKEN, awaitPromise: false });
  // 启用 Network 域，在 CDP 层面拦截 LLM 响应抓 token 用量
  await sessionObj.send("Network.enable");
 }
@@ -732,18 +1019,19 @@ async function connectToTarget(port) {
  closeNetworkSessions();
  state.target = { id: tgt.id, url: tgt.url, title: tgt.title };
  setPhase("connecting");
-  session = await openSession(tgt.webSocketDebuggerUrl, handleBinding);
-  session.on("Runtime.bindingCalled", handleBinding);
+  const ownedSession = await openSession(tgt.webSocketDebuggerUrl, handleBinding);
+  session = ownedSession;
  // Token 用量：Network 层拦截 LLM 响应
-  bindNetworkHandlers(session);
-  session.ws.on("close", () => {
+  bindNetworkHandlers(ownedSession);
+  ownedSession.ws.on("close", () => {
+    if (session !== ownedSession) return;
    stopTokenSync();
    closeNetworkSessions();
    session = null;
    setPhase("disconnected");
  });
  setPhase("injecting");
-  await inject(session);
+  await inject(ownedSession);
  startTokenSync();
  setPhase("connected", { target: state.target });
  attachWorkerNetworkTargets(port, tgt.id).catch((e) => {
@@ -755,44 +1043,66 @@ async function connectToTarget(port) {
 // —— 轮询直到出现 page target（重启后用）——
 async function waitForReady(port) {
  const t0 = Date.now();
-  while (Date.now() - t0 < READY_POLL_MAX_MS) {
+  const maxMs = state.codexPackaged ? PACKAGED_READY_POLL_MAX_MS : READY_POLL_MAX_MS;
+  let diagnosed = false;
+  while (Date.now() - t0 < maxMs) {
    const targets = await listTargets(port);
    const tgt = pickTarget(targets);
    if (tgt) return tgt;
+    if (!diagnosed && process.platform === "win32" && Date.now() - t0 >= 5000) {
+      diagnosed = true;
+      const d = updateWindowsProcessDiagnostic(port);
+      setPhase("waiting_debug_port", { portDiagnostic: state.portDiagnostic });
+      if (state.codexPackaged && d && Array.isArray(d.rows) && d.rows.length && !d.hasDebugArg) {
+        return null;
+      }
+    }
    await new Promise((r) => setTimeout(r, READY_POLL_INTERVAL));
  }
+  updateWindowsProcessDiagnostic(port);
  return null;
 }

-// —— 启动接管：杀当前 Codex → 带端口重启 ——
-function killCodex() {
-  return new Promise((resolve) => {
-    if (process.platform === "darwin") {
-      execFile("osascript", ["-e", 'tell application "ChatGPT" to quit'], () => {
-        // 兜底：osascript 没杀掉就 pkill
-        execFile("pgrep", ["-f", "ChatGPT.app/Contents/MacOS/ChatGPT"], (_e, out) => {
-          if (out && out.trim()) {
-            execFile("pkill", ["-f", "ChatGPT.app/Contents/MacOS/ChatGPT"], () =>
-              setTimeout(resolve, 800)
-            );
-          } else resolve();
-        });
-      });
-    } else if (process.platform === "win32") {
-      execFile("taskkill", ["/IM", "ChatGPT.exe", "/F"], () =>
-        setTimeout(resolve, 800)
-      );
-    } else resolve();
-  });
-}
 function spawnCodex(exePath, port) {
-  const child = spawn(exePath, [`--remote-debugging-port=${port}`], {
-    detached: true,
-    stdio: "ignore",
+  if (state.codexPackaged && process.platform === "win32") {
+    return windowsShellStart(exePath, port);
+  }
+  return new Promise((resolve, reject) => {
+    let child = null;
+    try {
+      const args = process.platform === "win32" ? windowsLaunchArgs(port) : [`--remote-debugging-port=${port}`];
+      child = spawn(exePath, args, {
+        detached: true,
+        stdio: "ignore",
+        windowsHide: false,
+      });
+    } catch (e) {
+      reject(e);
+      return;
+    }
+    state.launchWarning = "";
+    let settled = false;
+    const finish = (err) => {
+      if (settled) return;
+      settled = true;
+      if (err) {
+        reject(err);
+        return;
+      }
+      try {
+        child.unref();
+      } catch {}
+      resolve(true);
+    };
+    child.once("error", (err) => {
+      if (process.platform === "win32" && /eperm/i.test(String((err && err.message) || err || ""))) {
+        windowsShellStart(exePath, port).then(() => finish(null)).catch(finish);
+        return;
+      }
+      finish(err);
+    });
+    child.once("spawn", () => finish(null));
  });
-  try {
-    child.unref();
-  } catch {}
 }

 // —— 对外：status / connect / disconnect ——
@@ -802,31 +1112,39 @@ function status() {
 function onStatus(cb) {
  onStatusCb = cb;
 }
+function onSessionDelete(cb) {
+  onSessionDeleteCb = cb;
+}

 // connect(onStatusCb)：流程见计划。forceRestart=true 时直接重启（前端已确认）。
 async function connect(cb, opts = {}) {
  if (cb) onStatusCb = cb;
  const port = opts.port || DEFAULT_PORT;
  state.port = port;
-  setPhase("checking");
+  setPhase("checking", { lastError: "", launchWarning: "", portDiagnostic: "" });
+  const exe = scanCodexExecutable();
  // 1. 端口已开？
  const ok = await connectToTarget(port);
  if (ok) return snapshot();
  // 2. 判断 Codex 是否在跑
-  const running = providerSync.codexRunning ? providerSync.codexRunning() : false;
+  const running = codexDesktop.codexRunning();
  if (!running && !opts.forceRestart) {
    // 没在跑也没开端口：直接尝试拉起带端口的实例（无需杀进程）
-    const exe = codexExecutable();
    if (!exe) {
-      setPhase("error", { lastError: "未找到 Codex 可执行文件" });
+      setPhase("error", { lastError: desktopNotFoundError(), discovery: state.discovery });
      return snapshot();
    }
-    state.codexPath = exe;
    setPhase("restarting", { codexPath: exe });
-    spawnCodex(exe, port);
+    try {
+      await spawnCodex(exe, port);
+    } catch (e) {
+      setPhase("error", { lastError: desktopLaunchError(exe, e), discovery: state.discovery });
+      return snapshot();
+    }
+    setPhase("waiting_debug_port", { codexPath: exe, lastError: "" });
    const tgt = await waitForReady(port);
    if (!tgt) {
-      setPhase("error", { lastError: "Codex 启动后未开放调试端口，请重试" });
+      setDebugPortFailure("启动", port);
      return snapshot();
    }
    const ok2 = await connectToTarget(port);
@@ -835,23 +1153,27 @@ async function connect(cb, opts = {}) {
  }
  // 3. 在跑但没端口：需用户在前端确认后 forceRestart 重启
  if (running && !opts.forceRestart) {
-    state.codexPath = codexExecutable();
    setPhase("running_no_port", { codexPath: state.codexPath });
    return snapshot(); // 前端据此弹"重启确认"
  }
  // 4. forceRestart：杀进程 → 重启
-  const exe = codexExecutable();
  if (!exe) {
-    setPhase("error", { lastError: "未找到 Codex 可执行文件" });
+    setPhase("error", { lastError: desktopNotFoundError(), discovery: state.discovery });
    return snapshot();
  }
  state.codexPath = exe;
  setPhase("restarting");
-  await killCodex();
-  spawnCodex(exe, port);
+  await codexDesktop.killCodex();
+  try {
+    await spawnCodex(exe, port);
+  } catch (e) {
+    setPhase("error", { lastError: desktopLaunchError(exe, e), discovery: state.discovery });
+    return snapshot();
+  }
+  setPhase("waiting_debug_port", { codexPath: exe, lastError: "" });
  const tgt = await waitForReady(port);
  if (!tgt) {
-    setPhase("error", { lastError: "重启 Codex 后未开放调试端口" });
+    setDebugPortFailure("重启", port);
    return snapshot();
  }
  const ok3 = await connectToTarget(port);
@@ -859,14 +1181,28 @@ async function connect(cb, opts = {}) {
  return snapshot();
 }

-function disconnect() {
+async function disconnect() {
  stopTokenSync();
  closeNetworkSessions();
  if (session) {
+    const ownedSession = session;
+    // 先让页面停止录音并删除已挂载按钮，再关闭 CDP；关闭后无法清理前台 DOM。
    try {
-      session.close();
+      await ownedSession.send("Runtime.evaluate", {
+        expression: "window.__voiceInjectDispose && window.__voiceInjectDispose()",
+        awaitPromise: false,
+      });
    } catch {}
+    if (voiceScriptIdentifier) {
+      try {
+        await ownedSession.send("Page.removeScriptToEvaluateOnNewDocument", { identifier: voiceScriptIdentifier });
+      } catch {}
+      voiceScriptIdentifier = "";
+    }
    session = null;
+    try {
+      ownedSession.close();
+    } catch {}
  }
  setPhase("disconnected");
  return snapshot();
@@ -973,6 +1309,7 @@ module.exports = {
  connect,
  disconnect,
  onStatus,
+  onSessionDelete,
  sttInfo,
  probeStt,
  tokenStatus,