4
diff --git a/scripts/diagnose-windows.ps1 b/scripts/diagnose-windows.ps1
new file mode 100644
index 0000000..4dbe1b0
--- /dev/null
+++ b/scripts/diagnose-windows.ps1
@@ -0,0 +1,46 @@
+[CmdletBinding()]
+param([int]$Port = 0)
+
+$ErrorActionPreference = "Stop"
+$ProgressPreference = "SilentlyContinue"
+[Console]::OutputEncoding = New-Object System.Text.UTF8Encoding($false)
+
+if ($Port -eq 0) {
+ $parsedPort = 0
+ if ($env:CODEX_SESSION_GRID_PORT -and [int]::TryParse($env:CODEX_SESSION_GRID_PORT, [ref]$parsedPort)) {
+ $Port = $parsedPort
+ }
+ else {
+ $Port = 4317
+ }
+}
+
+$report = [ordered]@{
+ ok = $true
+ platform = "windows"
+ powershell = $PSVersionTable.PSVersion.ToString()
+ node = $null
+ codex = $null
+ health = $null
+ nativeBridge = $null
+}
+
+try { $report.node = (& node --version 2>&1 | Select-Object -First 1) } catch { $report.node = $_.Exception.Message }
+try { $report.codex = (& codex --version 2>&1 | Select-Object -First 1) } catch { $report.codex = $_.Exception.Message }
+try {
+ $report.health = Invoke-RestMethod -Uri "http://127.0.0.1:$Port/api/health" -TimeoutSec 2
+}
+catch {
+ $report.health = [ordered]@{ ok = $false; error = $_.Exception.Message }
+}
+
+$bridgeScript = Join-Path $PSScriptRoot "native-bridge-windows.ps1"
+try {
+ $bridgeOutput = & powershell.exe -NoLogo -NoProfile -NonInteractive -ExecutionPolicy Bypass -File $bridgeScript -Action probe
+ $report.nativeBridge = $bridgeOutput | ConvertFrom-Json
+}
+catch {
+ $report.nativeBridge = [ordered]@{ ok = $false; error = $_.Exception.Message }
+}
+
+$report | ConvertTo-Json -Depth 8
diff --git a/scripts/native-bridge-windows.cs b/scripts/native-bridge-windows.cs
new file mode 100644
index 0000000..cbd1e64
--- /dev/null
+++ b/scripts/native-bridge-windows.cs
@@ -0,0 +1,587 @@
+using System;
+using System.Collections.Generic;
+using System.ComponentModel;
+using System.Diagnostics;
+using System.Runtime.InteropServices;
+using System.Threading;
+using System.Windows;
+using System.Windows.Automation;
+
+namespace CodexSessionGrid
+{
+ public static class WindowsAutomationBridge
+ {
+ private const uint InputKeyboard = 1;
+ private const uint KeyEventKeyUp = 0x0002;
+ private const uint KeyEventUnicode = 0x0004;
+ private const ushort VkBack = 0x08;
+ private const ushort VkReturn = 0x0D;
+ private const ushort VkShift = 0x10;
+ private const ushort VkControl = 0x11;
+ private const ushort VkA = 0x41;
+ private const int RestoreWindow = 9;
+
+ private sealed class ComposerCandidate
+ {
+ public AutomationElement Element;
+ public int Score;
+ public string Label;
+ public Rect Frame;
+ }
+
+ private sealed class Target
+ {
+ public Process Process;
+ public AutomationElement Root;
+ public ComposerCandidate Composer;
+ public int CandidateCount;
+ }
+
+ [StructLayout(LayoutKind.Sequential)]
+ private struct MouseInput
+ {
+ public int Dx;
+ public int Dy;
+ public uint MouseData;
+ public uint Flags;
+ public uint Time;
+ public UIntPtr ExtraInfo;
+ }
+
+ [StructLayout(LayoutKind.Sequential)]
+ private struct KeyboardInput
+ {
+ public ushort VirtualKey;
+ public ushort ScanCode;
+ public uint Flags;
+ public uint Time;
+ public UIntPtr ExtraInfo;
+ }
+
+ [StructLayout(LayoutKind.Sequential)]
+ private struct HardwareInput
+ {
+ public uint Message;
+ public ushort ParameterLow;
+ public ushort ParameterHigh;
+ }
+
+ [StructLayout(LayoutKind.Explicit)]
+ private struct InputUnion
+ {
+ [FieldOffset(0)] public MouseInput Mouse;
+ [FieldOffset(0)] public KeyboardInput Keyboard;
+ [FieldOffset(0)] public HardwareInput Hardware;
+ }
+
+ [StructLayout(LayoutKind.Sequential)]
+ private struct Input
+ {
+ public uint Type;
+ public InputUnion Data;
+ }
+
+ [DllImport("user32.dll")]
+ private static extern IntPtr GetForegroundWindow();
+
+ [DllImport("user32.dll")]
+ [return: MarshalAs(UnmanagedType.Bool)]
+ private static extern bool SetForegroundWindow(IntPtr window);
+
+ [DllImport("user32.dll")]
+ [return: MarshalAs(UnmanagedType.Bool)]
+ private static extern bool ShowWindow(IntPtr window, int command);
+
+ [DllImport("user32.dll", SetLastError = true)]
+ private static extern uint SendInput(uint count, Input[] inputs, int size);
+
+ public static IDictionary
Probe()
+ {
+ Dictionary result = BaseResult();
+ List applications = FindApplicationProcesses();
+ result["appRunning"] = applications.Count > 0;
+ if (applications.Count == 0)
+ {
+ result["composerFound"] = false;
+ result["trusted"] = true;
+ return result;
+ }
+
+ Target target = FindBestTarget(applications);
+ result["composerFound"] = target != null && target.Composer != null;
+ result["trusted"] = target != null;
+ if (target == null)
+ {
+ result["error"] = "检测到 Codex/ChatGPT，但无法读取窗口。请确认工作台和桌面 App 使用相同权限级别。";
+ return result;
+ }
+
+ result["processName"] = SafeProcessName(target.Process);
+ result["candidateCount"] = target.CandidateCount;
+ if (target.Composer != null) result["composer"] = DescribeComposer(target.Composer);
+ else result["candidates"] = DescribeEditableCandidates(target.Root);
+ return result;
+ }
+
+ public static IDictionary Send(string threadId, string message)
+ {
+ Guid parsed;
+ if (!Guid.TryParse(threadId, out parsed)) throw new ArgumentException("Windows 原生桥收到的任务 ID 无效");
+ if (String.IsNullOrWhiteSpace(message)) throw new ArgumentException("发送内容不能为空");
+
+ IntPtr previousForeground = GetForegroundWindow();
+ OpenThread(threadId);
+
+ Target target = null;
+ for (int attempt = 0; attempt < 80; attempt++)
+ {
+ target = FindBestTarget(FindApplicationProcesses());
+ if (target != null && target.Composer != null) break;
+ Thread.Sleep(250);
+ }
+
+ if (target == null || target.Composer == null)
+ {
+ RestoreForeground(previousForeground);
+ throw new InvalidOperationException(
+ "无法识别 Windows Codex 输入框。请先打开 Codex 任务，并确保工作台与 Codex App 使用相同权限级别。"
+ );
+ }
+
+ string inputMethod = SetComposerValue(target, message);
+ Thread.Sleep(120);
+
+ string submissionMethod = InvokeSendButton(target);
+ if (submissionMethod == null)
+ {
+ FocusTarget(target);
+ PressVirtualKey(VkReturn);
+ submissionMethod = "keyboard-enter";
+ }
+
+ Thread.Sleep(160);
+ bool restored = RestoreForeground(previousForeground);
+ Dictionary result = BaseResult();
+ result["native"] = true;
+ result["threadId"] = threadId;
+ result["submitted"] = true;
+ result["inputMethod"] = inputMethod;
+ result["submissionMethod"] = submissionMethod;
+ result["foregroundRestored"] = restored;
+ result["processName"] = SafeProcessName(target.Process);
+ return result;
+ }
+
+ private static Dictionary BaseResult()
+ {
+ Dictionary result = new Dictionary();
+ result["ok"] = true;
+ result["supported"] = true;
+ result["platform"] = "windows";
+ result["permissionMode"] = "windows-uia";
+ return result;
+ }
+
+ private static IDictionary DescribeComposer(ComposerCandidate composer)
+ {
+ Dictionary description = new Dictionary();
+ description["label"] = composer.Label;
+ description["score"] = composer.Score;
+ description["x"] = composer.Frame.IsEmpty ? 0 : (int)composer.Frame.Left;
+ description["y"] = composer.Frame.IsEmpty ? 0 : (int)composer.Frame.Top;
+ description["width"] = composer.Frame.IsEmpty ? 0 : (int)composer.Frame.Width;
+ description["height"] = composer.Frame.IsEmpty ? 0 : (int)composer.Frame.Height;
+ description["controlType"] = SafeString(composer.Element, AutomationElement.ControlTypeProperty);
+ description["automationId"] = SafeString(composer.Element, AutomationElement.AutomationIdProperty);
+ description["className"] = SafeString(composer.Element, AutomationElement.ClassNameProperty);
+ return description;
+ }
+
+ private static IList> DescribeEditableCandidates(AutomationElement root)
+ {
+ List> descriptions = new List>();
+ Condition controls = new OrCondition(
+ new PropertyCondition(AutomationElement.ControlTypeProperty, ControlType.Edit),
+ new PropertyCondition(AutomationElement.ControlTypeProperty, ControlType.Document)
+ );
+ AutomationElementCollection elements = root.FindAll(TreeScope.Descendants, controls);
+ int count = Math.Min(20, elements.Count);
+ for (int index = 0; index < count; index++)
+ {
+ AutomationElement element = elements[index];
+ Rect frame = SafeRect(element);
+ Dictionary item = new Dictionary();
+ item["controlType"] = SafeString(element, AutomationElement.ControlTypeProperty);
+ item["automationId"] = SafeString(element, AutomationElement.AutomationIdProperty);
+ item["className"] = SafeString(element, AutomationElement.ClassNameProperty);
+ item["enabled"] = SafeBool(element, AutomationElement.IsEnabledProperty);
+ item["focusable"] = SafeBool(element, AutomationElement.IsKeyboardFocusableProperty);
+ item["valuePattern"] = SupportsPattern(element, ValuePattern.Pattern);
+ item["x"] = frame.IsEmpty ? 0 : (int)frame.Left;
+ item["y"] = frame.IsEmpty ? 0 : (int)frame.Top;
+ item["width"] = frame.IsEmpty ? 0 : (int)frame.Width;
+ item["height"] = frame.IsEmpty ? 0 : (int)frame.Height;
+ descriptions.Add(item);
+ }
+ return descriptions;
+ }
+
+ private static List FindApplicationProcesses()
+ {
+ List candidates = new List();
+ foreach (Process process in Process.GetProcesses())
+ {
+ try
+ {
+ if (process.MainWindowHandle == IntPtr.Zero) continue;
+ string name = process.ProcessName.ToLowerInvariant();
+ string title = (process.MainWindowTitle ?? String.Empty).ToLowerInvariant();
+ bool direct = name.Contains("codex") || name.Contains("chatgpt") || name.Contains("openai");
+ bool packaged = name == "applicationframehost" && (title.Contains("codex") || title.Contains("chatgpt"));
+ if (direct || packaged) candidates.Add(process);
+ }
+ catch
+ {
+ process.Dispose();
+ }
+ }
+ candidates.Sort(delegate(Process left, Process right)
+ {
+ return ApplicationScore(right).CompareTo(ApplicationScore(left));
+ });
+ return candidates;
+ }
+
+ private static int ApplicationScore(Process process)
+ {
+ try
+ {
+ string name = process.ProcessName.ToLowerInvariant();
+ string title = (process.MainWindowTitle ?? String.Empty).ToLowerInvariant();
+ int score = 0;
+ if (name == "codex" || name == "chatgpt") score += 120;
+ if (name.Contains("codex")) score += 80;
+ if (name.Contains("chatgpt")) score += 60;
+ if (name.Contains("openai")) score += 35;
+ if (title.Contains("codex")) score += 45;
+ if (title.Contains("chatgpt")) score += 25;
+ return score;
+ }
+ catch
+ {
+ return 0;
+ }
+ }
+
+ private static Target FindBestTarget(List applications)
+ {
+ Target best = null;
+ foreach (Process process in applications)
+ {
+ try
+ {
+ AutomationElement root = AutomationElement.FromHandle(process.MainWindowHandle);
+ if (root == null) continue;
+ int candidateCount;
+ ComposerCandidate composer = FindComposer(root, out candidateCount);
+ Target current = new Target();
+ current.Process = process;
+ current.Root = root;
+ current.Composer = composer;
+ current.CandidateCount = candidateCount;
+ if (best == null || (composer != null && (best.Composer == null || composer.Score > best.Composer.Score)))
+ {
+ best = current;
+ }
+ }
+ catch (ElementNotAvailableException) {}
+ catch (UnauthorizedAccessException) {}
+ catch (InvalidOperationException) {}
+ }
+ return best;
+ }
+
+ private static ComposerCandidate FindComposer(AutomationElement root, out int candidateCount)
+ {
+ Condition controls = new OrCondition(
+ new PropertyCondition(AutomationElement.ControlTypeProperty, ControlType.Edit),
+ new PropertyCondition(AutomationElement.ControlTypeProperty, ControlType.Document)
+ );
+ AutomationElementCollection elements = root.FindAll(TreeScope.Descendants, controls);
+ candidateCount = elements.Count;
+ Rect windowFrame = SafeRect(root);
+ ComposerCandidate best = null;
+ foreach (AutomationElement element in elements)
+ {
+ int score = ComposerScore(element, windowFrame);
+ if (score < 1) continue;
+ if (best == null || score > best.Score)
+ {
+ best = new ComposerCandidate();
+ best.Element = element;
+ best.Score = score;
+ best.Label = NormalizedLabel(element);
+ best.Frame = SafeRect(element);
+ }
+ }
+ return best;
+ }
+
+ private static int ComposerScore(AutomationElement element, Rect windowFrame)
+ {
+ string name = SafeString(element, AutomationElement.NameProperty).ToLowerInvariant();
+ string help = SafeString(element, AutomationElement.HelpTextProperty).ToLowerInvariant();
+ string automationId = SafeString(element, AutomationElement.AutomationIdProperty).ToLowerInvariant();
+ string className = SafeString(element, AutomationElement.ClassNameProperty).ToLowerInvariant();
+ string label = name + " " + help + " " + automationId + " " + className;
+ if (ContainsAny(label, "search", "filter", "find", "筛选", "搜索")) return -200;
+ if (SafeBool(element, AutomationElement.IsPasswordProperty)) return -200;
+
+ int score = 0;
+ ControlType type = SafeControlType(element);
+ if (type == ControlType.Edit) score += 35;
+ if (type == ControlType.Document) score += 8;
+ if (SafeBool(element, AutomationElement.IsEnabledProperty)) score += 15;
+ if (SafeBool(element, AutomationElement.IsKeyboardFocusableProperty)) score += 12;
+ if (ContainsAny(label, "composer", "prompt", "textarea")) score += 70;
+ if (ContainsAny(label, "send a message", "message codex", "ask codex", "发送消息", "向这个任务发送消息", "输入消息")) score += 80;
+ if (SupportsPattern(element, ValuePattern.Pattern)) score += 30;
+
+ Rect frame = SafeRect(element);
+ if (!frame.IsEmpty)
+ {
+ if (frame.Width >= 220) score += 18;
+ if (frame.Height >= 24 && frame.Height <= 360) score += 12;
+ if (!windowFrame.IsEmpty && frame.Top >= windowFrame.Top + windowFrame.Height * 0.5) score += 35;
+ }
+ return score;
+ }
+
+ private static string SetComposerValue(Target target, string message)
+ {
+ object pattern;
+ if (target.Composer.Element.TryGetCurrentPattern(ValuePattern.Pattern, out pattern))
+ {
+ ValuePattern value = pattern as ValuePattern;
+ if (value != null && !value.Current.IsReadOnly)
+ {
+ value.SetValue(message);
+ return "value-pattern";
+ }
+ }
+
+ FocusTarget(target);
+ PressChord(VkControl, VkA);
+ PressVirtualKey(VkBack);
+ TypeUnicode(message);
+ return "keyboard-unicode";
+ }
+
+ private static string InvokeSendButton(Target target)
+ {
+ Condition buttonsOnly = new PropertyCondition(AutomationElement.ControlTypeProperty, ControlType.Button);
+ AutomationElementCollection buttons = target.Root.FindAll(TreeScope.Descendants, buttonsOnly);
+ AutomationElement best = null;
+ int bestScore = Int32.MinValue;
+ Rect composerFrame = target.Composer.Frame;
+ foreach (AutomationElement button in buttons)
+ {
+ if (!SafeBool(button, AutomationElement.IsEnabledProperty)) continue;
+ string label = NormalizedLabel(button).ToLowerInvariant();
+ if (!ContainsAny(label, "send", "submit", "发送", "提交")) continue;
+ int score = 100;
+ Rect frame = SafeRect(button);
+ if (!frame.IsEmpty && !composerFrame.IsEmpty)
+ {
+ double dx = frame.Left + frame.Width / 2 - (composerFrame.Left + composerFrame.Width / 2);
+ double dy = frame.Top + frame.Height / 2 - (composerFrame.Top + composerFrame.Height / 2);
+ score -= (int)Math.Min(80, Math.Sqrt(dx * dx + dy * dy) / 20);
+ }
+ if (score > bestScore)
+ {
+ best = button;
+ bestScore = score;
+ }
+ }
+
+ if (best == null) return null;
+ object pattern;
+ if (!best.TryGetCurrentPattern(InvokePattern.Pattern, out pattern)) return null;
+ InvokePattern invoke = pattern as InvokePattern;
+ if (invoke == null) return null;
+ invoke.Invoke();
+ return "uia-invoke";
+ }
+
+ private static void OpenThread(string threadId)
+ {
+ ProcessStartInfo start = new ProcessStartInfo();
+ start.FileName = "codex://threads/" + threadId;
+ start.UseShellExecute = true;
+ try
+ {
+ Process process = Process.Start(start);
+ if (process != null) process.Dispose();
+ }
+ catch (Exception error)
+ {
+ throw new InvalidOperationException("无法通过 codex:// 链接打开目标任务：" + error.Message, error);
+ }
+ }
+
+ private static void FocusTarget(Target target)
+ {
+ IntPtr window = target.Process.MainWindowHandle;
+ if (window != IntPtr.Zero)
+ {
+ ShowWindow(window, RestoreWindow);
+ SetForegroundWindow(window);
+ }
+ target.Composer.Element.SetFocus();
+ Thread.Sleep(80);
+ }
+
+ private static bool RestoreForeground(IntPtr previous)
+ {
+ if (previous == IntPtr.Zero) return false;
+ ShowWindow(previous, RestoreWindow);
+ SetForegroundWindow(previous);
+ Thread.Sleep(80);
+ return GetForegroundWindow() == previous;
+ }
+
+ private static void TypeUnicode(string value)
+ {
+ List inputs = new List();
+ foreach (char character in value)
+ {
+ if (character == '\r') continue;
+ if (character == '\n')
+ {
+ FlushInputs(inputs);
+ PressChord(VkShift, VkReturn);
+ continue;
+ }
+ inputs.Add(UnicodeInput(character, false));
+ inputs.Add(UnicodeInput(character, true));
+ if (inputs.Count >= 256) FlushInputs(inputs);
+ }
+ FlushInputs(inputs);
+ }
+
+ private static Input UnicodeInput(char character, bool keyUp)
+ {
+ Input input = new Input();
+ input.Type = InputKeyboard;
+ input.Data.Keyboard = new KeyboardInput();
+ input.Data.Keyboard.ScanCode = character;
+ input.Data.Keyboard.Flags = KeyEventUnicode | (keyUp ? KeyEventKeyUp : 0);
+ return input;
+ }
+
+ private static Input VirtualKeyInput(ushort key, bool keyUp)
+ {
+ Input input = new Input();
+ input.Type = InputKeyboard;
+ input.Data.Keyboard = new KeyboardInput();
+ input.Data.Keyboard.VirtualKey = key;
+ input.Data.Keyboard.Flags = keyUp ? KeyEventKeyUp : 0;
+ return input;
+ }
+
+ private static void PressVirtualKey(ushort key)
+ {
+ SendInputs(new Input[] { VirtualKeyInput(key, false), VirtualKeyInput(key, true) });
+ }
+
+ private static void PressChord(ushort modifier, ushort key)
+ {
+ SendInputs(new Input[]
+ {
+ VirtualKeyInput(modifier, false),
+ VirtualKeyInput(key, false),
+ VirtualKeyInput(key, true),
+ VirtualKeyInput(modifier, true),
+ });
+ }
+
+ private static void FlushInputs(List inputs)
+ {
+ if (inputs.Count == 0) return;
+ SendInputs(inputs.ToArray());
+ inputs.Clear();
+ }
+
+ private static void SendInputs(Input[] inputs)
+ {
+ uint sent = SendInput((uint)inputs.Length, inputs, Marshal.SizeOf(typeof(Input)));
+ if (sent != (uint)inputs.Length) throw new Win32Exception(Marshal.GetLastWin32Error(), "Windows 无法向 Codex 输入文字");
+ }
+
+ private static string NormalizedLabel(AutomationElement element)
+ {
+ return String.Join(" ", new string[]
+ {
+ SafeString(element, AutomationElement.NameProperty),
+ SafeString(element, AutomationElement.HelpTextProperty),
+ SafeString(element, AutomationElement.AutomationIdProperty),
+ SafeString(element, AutomationElement.ClassNameProperty),
+ }).Trim();
+ }
+
+ private static bool ContainsAny(string text, params string[] values)
+ {
+ foreach (string value in values)
+ {
+ if (text.IndexOf(value, StringComparison.OrdinalIgnoreCase) >= 0) return true;
+ }
+ return false;
+ }
+
+ private static bool SupportsPattern(AutomationElement element, AutomationPattern pattern)
+ {
+ object value;
+ try { return element.TryGetCurrentPattern(pattern, out value); }
+ catch { return false; }
+ }
+
+ private static string SafeString(AutomationElement element, AutomationProperty property)
+ {
+ try
+ {
+ object value = element.GetCurrentPropertyValue(property, true);
+ if (value == null || value == AutomationElement.NotSupported) return String.Empty;
+ return Convert.ToString(value) ?? String.Empty;
+ }
+ catch { return String.Empty; }
+ }
+
+ private static bool SafeBool(AutomationElement element, AutomationProperty property)
+ {
+ try
+ {
+ object value = element.GetCurrentPropertyValue(property, true);
+ return value is bool && (bool)value;
+ }
+ catch { return false; }
+ }
+
+ private static Rect SafeRect(AutomationElement element)
+ {
+ try { return element.Current.BoundingRectangle; }
+ catch { return Rect.Empty; }
+ }
+
+ private static ControlType SafeControlType(AutomationElement element)
+ {
+ try { return element.Current.ControlType; }
+ catch { return null; }
+ }
+
+ private static string SafeProcessName(Process process)
+ {
+ try { return process.ProcessName; }
+ catch { return "unknown"; }
+ }
+ }
+}
diff --git a/scripts/native-bridge-windows.ps1 b/scripts/native-bridge-windows.ps1
new file mode 100644
index 0000000..0f0f723
--- /dev/null
+++ b/scripts/native-bridge-windows.ps1
@@ -0,0 +1,51 @@
+param(
+ [ValidateSet("probe", "send")]
+ [string]$Action = "probe",
+ [string]$ThreadId = ""
+)
+
+$ErrorActionPreference = "Stop"
+$ProgressPreference = "SilentlyContinue"
+[Console]::OutputEncoding = New-Object System.Text.UTF8Encoding($false)
+[Console]::InputEncoding = New-Object System.Text.UTF8Encoding($false)
+
+function Write-BridgeJson {
+ param([object]$Value)
+ [Console]::Out.WriteLine(($Value | ConvertTo-Json -Depth 8 -Compress))
+}
+
+try {
+ if ($env:OS -ne "Windows_NT") {
+ throw "Windows UI Automation 桥只能在 Windows 上运行"
+ }
+
+ Add-Type -AssemblyName WindowsBase
+ Add-Type -AssemblyName UIAutomationClient
+ Add-Type -AssemblyName UIAutomationTypes
+
+ $sourcePath = Join-Path $PSScriptRoot "native-bridge-windows.cs"
+ $references = @(
+ [System.Windows.Rect].Assembly.Location,
+ [System.Windows.Automation.AutomationElement].Assembly.Location,
+ [System.Windows.Automation.AutomationIdentifier].Assembly.Location
+ ) | Select-Object -Unique
+ Add-Type -Path $sourcePath -ReferencedAssemblies $references
+
+ if ($Action -eq "probe") {
+ Write-BridgeJson ([CodexSessionGrid.WindowsAutomationBridge]::Probe())
+ exit 0
+ }
+
+ $message = [Console]::In.ReadToEnd()
+ Write-BridgeJson ([CodexSessionGrid.WindowsAutomationBridge]::Send($ThreadId, $message))
+ exit 0
+}
+catch {
+ Write-BridgeJson ([ordered]@{
+ ok = $false
+ platform = "windows"
+ supported = $true
+ error = $_.Exception.Message
+ })
+ exit 0
+}
diff --git a/scripts/server.mjs b/scripts/server.mjs
index a3ab933..8aab26f 100755
--- a/scripts/server.mjs
+++ b/scripts/server.mjs
@@ -14,8 +14,11 @@ const args = process.argv.slice(2);
const portArg = args.indexOf("--port");
const port = Number(portArg >= 0 ? args[portArg + 1] : process.env.CODEX_SESSION_GRID_PORT || 4317);
const host = "127.0.0.1";
-const nativeBridgeSource = join(scriptDir, "native-bridge.swift");
-const nativeBridgeBinary = join(tmpdir(), `codex-session-grid-native-bridge-${process.getuid?.() ?? "user"}`);
+const appVersion = "0.2.0-beta.1";
+const macBridgeSource = join(scriptDir, "native-bridge.swift");
+const macBridgeBinary = join(tmpdir(), `codex-session-grid-native-bridge-${process.getuid?.() ?? "user"}`);
+const windowsBridgeScript = join(scriptDir, "native-bridge-windows.ps1");
+const windowsBridgeSource = join(scriptDir, "native-bridge-windows.cs");
if (!Number.isInteger(port) || port < 1 || port > 65535) {
console.error("Invalid port. Use --port 4317 or set CODEX_SESSION_GRID_PORT.");
@@ -30,9 +33,9 @@ const mimeTypes = {
".svg": "image/svg+xml",
};
-function captureProcess(command, commandArgs, timeoutMs = 30_000) {
+function captureProcess(command, commandArgs, timeoutMs = 30_000, input = "") {
return new Promise((resolve, reject) => {
- const child = spawn(command, commandArgs, { stdio: ["ignore", "pipe", "pipe"], env: process.env });
+ const child = spawn(command, commandArgs, { stdio: ["pipe", "pipe", "pipe"], env: process.env });
let stdout = "";
let stderr = "";
const timer = setTimeout(() => {
@@ -53,6 +56,8 @@ function captureProcess(command, commandArgs, timeoutMs = 30_000) {
if (code === 0) resolve({ stdout, stderr });
else reject(new Error(stderr.trim() || stdout.trim() || `Native Codex bridge exited with ${code}`));
});
+ child.stdin.on("error", () => {});
+ child.stdin.end(input);
});
}
@@ -61,39 +66,86 @@ class NativeCodexBridge {
this.buildPromise = null;
}
+ get platform() {
+ if (process.platform === "darwin") return "macos";
+ if (process.platform === "win32") return "windows";
+ return process.platform;
+ }
+
+ get supported() {
+ return process.platform === "darwin" || process.platform === "win32";
+ }
+
async ensureBuilt() {
- if (process.platform !== "darwin") throw new Error("原生同步发送目前仅支持 macOS Codex App");
+ if (process.platform === "win32") {
+ await Promise.all([stat(windowsBridgeScript), stat(windowsBridgeSource)]);
+ return;
+ }
+ if (process.platform !== "darwin") throw new Error(`原生同步发送暂不支持 ${process.platform}`);
let rebuild = true;
try {
- const [sourceInfo, binaryInfo] = await Promise.all([stat(nativeBridgeSource), stat(nativeBridgeBinary)]);
+ const [sourceInfo, binaryInfo] = await Promise.all([stat(macBridgeSource), stat(macBridgeBinary)]);
rebuild = binaryInfo.mtimeMs < sourceInfo.mtimeMs;
} catch {}
if (!rebuild) return;
if (!this.buildPromise) {
- this.buildPromise = captureProcess("/usr/bin/swiftc", ["-O", "-o", nativeBridgeBinary, nativeBridgeSource], 60_000)
+ this.buildPromise = captureProcess("/usr/bin/swiftc", ["-O", "-o", macBridgeBinary, macBridgeSource], 60_000)
.finally(() => { this.buildPromise = null; });
}
await this.buildPromise;
}
- async run(commandArgs) {
+ async run(commandArgs, input = "") {
await this.ensureBuilt();
- const { stdout } = await captureProcess(nativeBridgeBinary, commandArgs, 20_000);
+ const powershell = process.env.CODEX_SESSION_GRID_POWERSHELL || "powershell.exe";
+ const windowsArgs = commandArgs[0] === "send"
+ ? ["-Action", "send", "-ThreadId", commandArgs[1]]
+ : ["-Action", "probe"];
+ const invocation = process.platform === "win32"
+ ? {
+ command: powershell,
+ args: [
+ "-NoLogo",
+ "-NoProfile",
+ "-NonInteractive",
+ "-ExecutionPolicy",
+ "Bypass",
+ "-File",
+ windowsBridgeScript,
+ ...windowsArgs,
+ ],
+ timeout: 30_000,
+ }
+ : { command: macBridgeBinary, args: commandArgs, timeout: 20_000 };
+ const { stdout } = await captureProcess(invocation.command, invocation.args, invocation.timeout, input);
let result;
try {
- result = JSON.parse(stdout);
+ result = JSON.parse(stdout.trim());
} catch {
throw new Error("原生同步桥返回了无法识别的结果");
}
if (!result.ok) throw new Error(result.error || "原生同步桥执行失败");
- return result;
+ return { ...result, platform: this.platform, supported: true };
}
probe() {
+ if (!this.supported) {
+ return Promise.resolve({
+ ok: true,
+ platform: this.platform,
+ supported: false,
+ trusted: false,
+ appRunning: false,
+ composerFound: false,
+ error: `原生同步发送暂不支持 ${process.platform}`,
+ });
+ }
return this.run(["probe"]);
}
send(threadId, text) {
+ if (!this.supported) throw new Error(`原生同步发送暂不支持 ${process.platform}`);
+ if (process.platform === "win32") return this.run(["send", threadId], text);
return this.run(["send", threadId, text]);
}
}
@@ -143,7 +195,7 @@ class CodexAppServer {
clientInfo: {
name: "codex-session-grid",
title: "Codex Session Grid",
- version: "0.1.0-beta.1",
+ version: appVersion,
},
capabilities: {
experimentalApi: true,
@@ -436,7 +488,7 @@ const server = createServer(async (request, response) => {
try {
if (request.method === "GET" && url.pathname === "/api/health") {
await bridge.ready;
- return sendJson(response, 200, { ok: true, version: "0.1.0-beta.1" });
+ return sendJson(response, 200, { ok: true, version: appVersion, platform: nativeBridge.platform });
}
if (request.method === "GET" && url.pathname === "/api/events") {
diff --git a/scripts/start.ps1 b/scripts/start.ps1
new file mode 100644
index 0000000..9593399
--- /dev/null
+++ b/scripts/start.ps1
@@ -0,0 +1,85 @@
+[CmdletBinding()]
+param(
+ [int]$Port = 0,
+ [switch]$NoOpen
+)
+
+$ErrorActionPreference = "Stop"
+$ProgressPreference = "SilentlyContinue"
+
+if ($Port -eq 0) {
+ $parsedPort = 0
+ if ($env:CODEX_SESSION_GRID_PORT -and [int]::TryParse($env:CODEX_SESSION_GRID_PORT, [ref]$parsedPort)) {
+ $Port = $parsedPort
+ }
+ else {
+ $Port = 4317
+ }
+}
+if ($Port -lt 1 -or $Port -gt 65535) {
+ throw "端口必须在 1 到 65535 之间"
+}
+
+$userToken = if ($env:USERNAME) { $env:USERNAME } else { "user" }
+$runtimeDir = Join-Path ([IO.Path]::GetTempPath()) "codex-session-grid-$userToken-$Port"
+$pidFile = Join-Path $runtimeDir "server.pid"
+$stdoutLog = Join-Path $runtimeDir "server.out.log"
+$stderrLog = Join-Path $runtimeDir "server.err.log"
+$gridUrl = "http://127.0.0.1:$Port"
+$serverPath = (Resolve-Path (Join-Path $PSScriptRoot "server.mjs")).Path
+
+New-Item -ItemType Directory -Force -Path $runtimeDir | Out-Null
+
+function Test-GridHealth {
+ try {
+ $health = Invoke-RestMethod -Uri "$gridUrl/api/health" -TimeoutSec 1
+ return $health.ok -eq $true
+ }
+ catch {
+ return $false
+ }
+}
+
+if (Test-GridHealth) {
+ Write-Host "Codex Session Grid 已在运行：$gridUrl"
+}
+else {
+ $node = Get-Command node.exe -ErrorAction SilentlyContinue
+ if (-not $node) { $node = Get-Command node -ErrorAction Stop }
+
+ Remove-Item $stdoutLog, $stderrLog -Force -ErrorAction SilentlyContinue
+ $serverArguments = @("`"$serverPath`"", "--port", "$Port")
+ $startParameters = @{
+ FilePath = $node.Source
+ ArgumentList = $serverArguments
+ WindowStyle = "Hidden"
+ RedirectStandardOutput = $stdoutLog
+ RedirectStandardError = $stderrLog
+ PassThru = $true
+ }
+ $server = Start-Process @startParameters
+ Set-Content -Path $pidFile -Value $server.Id -Encoding ASCII
+
+ $started = $false
+ for ($attempt = 0; $attempt -lt 600; $attempt += 1) {
+ if (Test-GridHealth) {
+ $started = $true
+ break
+ }
+ if ($server.HasExited) {
+ $errorText = if (Test-Path $stderrLog) { Get-Content $stderrLog -Raw } else { "" }
+ throw "Codex Session Grid 启动失败。$errorText"
+ }
+ Start-Sleep -Milliseconds 100
+ }
+ if (-not $started) {
+ Stop-Process -Id $server.Id -Force -ErrorAction SilentlyContinue
+ throw "Codex Session Grid 启动超时。日志：$stderrLog"
+ }
+ Write-Host "Codex Session Grid 已启动：$gridUrl"
+}
+
+$skipOpen = $NoOpen -or $env:CODEX_SESSION_GRID_NO_OPEN -eq "1"
+if (-not $skipOpen) {
+ Start-Process $gridUrl
+}
diff --git a/scripts/stop.ps1 b/scripts/stop.ps1
new file mode 100644
index 0000000..3034df6
--- /dev/null
+++ b/scripts/stop.ps1
@@ -0,0 +1,52 @@
+[CmdletBinding()]
+param([int]$Port = 0)
+
+$ErrorActionPreference = "Stop"
+
+if ($Port -eq 0) {
+ $parsedPort = 0
+ if ($env:CODEX_SESSION_GRID_PORT -and [int]::TryParse($env:CODEX_SESSION_GRID_PORT, [ref]$parsedPort)) {
+ $Port = $parsedPort
+ }
+ else {
+ $Port = 4317
+ }
+}
+if ($Port -lt 1 -or $Port -gt 65535) {
+ throw "端口必须在 1 到 65535 之间"
+}
+
+$userToken = if ($env:USERNAME) { $env:USERNAME } else { "user" }
+$runtimeDir = Join-Path ([IO.Path]::GetTempPath()) "codex-session-grid-$userToken-$Port"
+$pidFile = Join-Path $runtimeDir "server.pid"
+
+if (-not (Test-Path $pidFile)) {
+ Write-Host "端口 $Port 上没有由脚本记录的 Codex Session Grid。"
+ exit 0
+}
+
+$serverPid = 0
+$pidText = (Get-Content $pidFile -Raw).Trim()
+if (-not [int]::TryParse($pidText, [ref]$serverPid) -or $serverPid -lt 1) {
+ throw "PID 文件无效：$pidFile"
+}
+
+$process = Get-CimInstance Win32_Process -Filter "ProcessId = $serverPid" -ErrorAction SilentlyContinue
+if (-not $process) {
+ Remove-Item $pidFile -Force
+ Write-Host "Codex Session Grid 已经停止。"
+ exit 0
+}
+
+$expectedServer = (Resolve-Path (Join-Path $PSScriptRoot "server.mjs")).Path
+$commandLine = [string]$process.CommandLine
+if (-not $commandLine.Contains($expectedServer) -or -not $commandLine.Contains("--port")) {
+ throw "PID $serverPid 不是当前项目的 Codex Session Grid，已拒绝停止。"
+}
+
+& taskkill.exe /PID $serverPid /T /F | Out-Null
+if ($LASTEXITCODE -ne 0) {
+ throw "无法停止 Codex Session Grid（PID $serverPid）"
+}
+Remove-Item $pidFile -Force
+Write-Host "Codex Session Grid 已停止（PID $serverPid）。"
diff --git a/skills/codex-session-grid/SKILL.md b/skills/codex-session-grid/SKILL.md
index f5e0a05..fdbd307 100644
--- a/skills/codex-session-grid/SKILL.md
+++ b/skills/codex-session-grid/SKILL.md
@@ -1,11 +1,11 @@
---
name: codex-session-grid
-description: Start and manage a local Codex-styled multi-session dashboard with 2x2 and 3x3 layouts, voice input, guided prompts, and per-session model, speed, and reasoning controls. Use when the user asks to multi-open Codex sessions, view several tasks at once, open a four-grid or nine-grid workspace, monitor parallel Codex work, control session models, or launch/stop the Codex Session Grid sidecar.
+description: Start and manage a cross-platform local Codex-styled multi-session dashboard with 2x2 and 3x3 layouts, native single-send bridges for macOS and Windows, voice input, guided prompts, and model controls. Use when the user asks to multi-open Codex sessions, view several tasks at once, open a four-grid or nine-grid workspace, monitor parallel Codex work, or launch/stop the Codex Session Grid sidecar.
---
# Codex Session Grid
-Launch a localhost-only sidecar that talks to the bundled Codex app-server. It uses the existing Codex login and dynamically reads the installed model catalog; never ask for an API key. Messages sent from the grid use the macOS Accessibility bridge to submit through the native Codex App composer exactly once; the sidecar must never create a second `turn/start` for the same message.
+Launch a localhost-only sidecar that talks to the bundled Codex app-server. It uses the existing Codex login and dynamically reads the installed model catalog; never ask for an API key. Messages sent from the grid use the macOS Accessibility bridge or Windows UI Automation bridge to submit through the native Codex App composer exactly once; the sidecar must never create a second `turn/start` for the same message.
## Launch
@@ -15,11 +15,17 @@ Run the bundled launcher from this skill directory:
bash ../../scripts/start.sh
```
-Report the printed URL. The launcher reuses an existing healthy instance and opens it in the default browser on macOS or Linux.
+On Windows PowerShell, run:
+
+```powershell
+powershell.exe -NoProfile -ExecutionPolicy Bypass -File ../../scripts/start.ps1
+```
+
+Report the printed URL. The launcher reuses an existing healthy instance and opens it in the default browser.
When reporting the URL, include this first-use guidance: voice input requires a browser with Web Speech API support and microphone permission. After installing or updating the plugin, recommend starting a new Codex task so the latest skill is loaded. The same guidance is available from the help button in the workspace.
-Also explain that native single-send requires macOS Accessibility permission for Codex under System Settings → Privacy & Security → Accessibility. Sending opens and submits the native target task in the background, keeps the grid visible, and immediately shows an optimistic sent state while the same Session is synced. The native Codex App's current model, speed, and reasoning settings are authoritative for that send, so change them in the native composer when exact overrides are required.
+Also explain that macOS native single-send requires Accessibility permission for Codex under System Settings → Privacy & Security → Accessibility. On Windows, launch with PowerShell and keep the grid process at the same privilege level as Codex; if Codex runs as administrator, the launcher must also run as administrator. Sending opens and submits the native target task, restores the previous foreground window when possible, and immediately shows an optimistic sent state while the same Session is synced. The native Codex App's current model, speed, and reasoning settings are authoritative for that send, so change them in the native composer when exact overrides are required.
Use a different port only when requested or when the default port is occupied:
@@ -27,6 +33,8 @@ Use a different port only when requested or when the default port is occupied:
CODEX_SESSION_GRID_PORT=4318 bash ../../scripts/start.sh
```
+On Windows, pass `-Port 4318` to `start.ps1` and `stop.ps1`.
+
## Stop
Stop the default instance with:
@@ -35,6 +43,12 @@ Stop the default instance with:
bash ../../scripts/stop.sh
```
+On Windows PowerShell, run:
+
+```powershell
+powershell.exe -NoProfile -ExecutionPolicy Bypass -File ../../scripts/stop.ps1
+```
+
Pass the same `CODEX_SESSION_GRID_PORT` value when stopping a non-default instance.
## Guardrails
@@ -43,4 +57,5 @@ Pass the same `CODEX_SESSION_GRID_PORT` value when stopping a non-default instan
- Do not add authentication tokens or modify the user's Codex config.
- Treat the app-server API as experimental. If startup fails after a Codex update, run the launcher in the foreground with `node ../../scripts/server.mjs --port 4317` and report the protocol error.
- Do not claim that this plugin changes the native Codex desktop layout. It provides a companion workspace and deep-links individual tasks back to Codex.
-- Never reintroduce a sidecar `turn/start` as a fallback for web sends. If the native Accessibility bridge cannot find the composer, show the error and keep the draft so the user can retry without duplicate execution.
+- Never reintroduce a sidecar `turn/start` as a fallback for web sends. If either native bridge cannot find the composer, show the error and keep the draft so the user can retry without duplicate execution.
+- On Windows diagnostics, report only platform, tool versions, health, process name, element counts, and composer metadata. Do not emit Session messages or the user's draft.