deepseek-computer-use 0.1.2 → 0.2.0

package/native/Program.cs

@@ -1,163 +1,163 @@
-
-using System;
-using System.Collections.Generic;
-using System.Globalization;
-using System.IO;
-using System.Text;
-using System.Threading;
-
-// ---------------------------------------------------------------------------
-// stdio JSON-lines bridge: one request object per line in, one response object
-// per line out. The DSH plugin owns the process lifetime; this helper owns the
-// UI Automation / SendInput / capture state so element indexes stay valid.
-// ---------------------------------------------------------------------------
-public static class Program
-{
-    static IPlatformDriver _driver;
-    /// Set by the close RPC: the loop stops after flushing the response.
-    public static bool CloseRequested;
-    /// 130 mirrors Codex's "the user pressed Esc" exit code.
-    public static int RequestedExitCode;
-
-    public static int Main(string[] args)
-    {
-        try
-        {
-            Console.OutputEncoding = new UTF8Encoding(false);
-            Console.InputEncoding = new UTF8Encoding(false);
-        }
-        catch { }
-
-        _driver = CreateDriver();
-
-        if (args.Length > 0 && args[0] == "--self-test")
-        {
-            return SelfTest();
-        }
-
-        string line;
-        while (true)
-        {
-            try { line = Console.ReadLine(); }
-            catch (Exception) { break; }
-            if (line == null) break;
-            if (line.Trim().Length == 0) continue;
-            string response = Handle(line);
-            Console.Out.Write(response);
-            Console.Out.Write('\n');
+
+using System;
+using System.Collections.Generic;
+using System.Globalization;
+using System.IO;
+using System.Text;
+using System.Threading;
+
+// ---------------------------------------------------------------------------
+// stdio JSON-lines bridge: one request object per line in, one response object
+// per line out. The DSH plugin owns the process lifetime; this helper owns the
+// UI Automation / SendInput / capture state so element indexes stay valid.
+// ---------------------------------------------------------------------------
+public static class Program
+{
+    static IPlatformDriver _driver;
+    /// Set by the close RPC: the loop stops after flushing the response.
+    public static bool CloseRequested;
+    /// 130 mirrors Codex's "the user pressed Esc" exit code.
+    public static int RequestedExitCode;
+
+    public static int Main(string[] args)
+    {
+        try
+        {
+            Console.OutputEncoding = new UTF8Encoding(false);
+            Console.InputEncoding = new UTF8Encoding(false);
+        }
+        catch { }
+
+        _driver = CreateDriver();
+
+        if (args.Length > 0 && args[0] == "--self-test")
+        {
+            return SelfTest();
+        }
+
+        string line;
+        while (true)
+        {
+            try { line = Console.ReadLine(); }
+            catch (Exception) { break; }
+            if (line == null) break;
+            if (line.Trim().Length == 0) continue;
+            string response = Handle(line);
+            Console.Out.Write(response);
+            Console.Out.Write('\n');
             Console.Out.Flush();
             if (CloseRequested || RequestedExitCode != 0) break;
         }
-        try { if (_driver != null) _driver.Dispose(); } catch { }
-        return RequestedExitCode;
-    }
-
-    static IPlatformDriver CreateDriver()
-    {
-        string forced = Environment.GetEnvironmentVariable("DSH_CU_PLATFORM");
-        if (!string.IsNullOrEmpty(forced) && forced != "windows")
-            throw new NotSupportedException("platform '" + forced + "' has no driver implementation yet; only windows is available");
-        return new WindowsDriver();
-    }
-
-    static int SelfTest()
-    {
-        Dictionary<string, object> ping = _driver.Ping(new Dictionary<string, object>());
-        Dictionary<string, object> wins = _driver.ListWindows(new Dictionary<string, object>());
-        Dictionary<string, object> report = new Dictionary<string, object>();
-        report["ok"] = true;
-        report["platform"] = _driver.PlatformName;
-        report["windowCount"] = wins["count"];
-        report["driver"] = ping;
-        Console.Out.Write(Json.ToJson(report));
-        Console.Out.Write('\n');
-        Console.Out.Flush();
-        return 0;
-    }
-
-    static string Handle(string line)
-    {
-        long id = 0;
-        try
-        {
-            Dictionary<string, object> req = Json.AsObject(Json.Parse(line));
-            if (req == null) throw new ToolError("bad_request", "request must be a JSON object");
-            id = Json.GetLong(req, "id", 0);
-            string method = Json.GetString(req, "method");
-            Dictionary<string, object> p = Json.AsObject(Json.Get(req, "params"));
-            if (p == null) p = new Dictionary<string, object>();
-            // Codex ships the budget with every call (x-oai-cua-request-budget-ms) and the
-            // helper honours it; here it bounds capture waits and marks overruns.
-            Budget.Begin(Json.GetLong(p, "__budgetMs", 0), Json.GetString(p, "__cancelFile"));
-            try { if (_driver != null) _driver.Maintenance(); } catch { }
-            object result = Dispatch(method, p);
-            return Ok(id, result);
-        }
-        catch (ToolError te)
-        {
-            try { _driver.InvalidateState(new Dictionary<string, object>()); } catch { }
-            return Fail(id, te.Code, te.Message);
-        }
-        catch (Exception ex)
-        {
-            try { _driver.InvalidateState(new Dictionary<string, object>()); } catch { }
-            return Fail(id, "internal_error", ex.GetType().Name + ": " + ex.Message);
-        }
-    }
-
-    static object Dispatch(string method, Dictionary<string, object> p)
-    {
-        if (string.IsNullOrEmpty(method)) throw new ToolError("bad_request", "request.method is required");
-        switch (method)
-        {
-            case "ping": return _driver.Ping(p);
-            case "hello": return _driver.Hello(p);
-            case "restore_cursor": return _driver.RestoreCursor(p);
-            case "close": return _driver.Close(p);
-            case "list_windows": return _driver.ListWindows(p);
-            case "list_apps": return _driver.ListApps(p);
-            case "get_window": return _driver.GetWindow(p);
-            case "launch_app": return _driver.LaunchApp(p);
-            case "get_window_state": return _driver.GetWindowState(p);
-            case "click": return _driver.Click(p);
-            case "press_key": return _driver.PressKey(p);
-            case "type_text": return _driver.TypeText(p);
-            case "scroll": return _driver.Scroll(p);
-            case "drag": return _driver.Drag(p);
-            case "set_value": return _driver.SetValue(p);
-            case "perform_secondary_action": return _driver.PerformSecondaryAction(p);
-            case "activate_window": return _driver.ActivateWindow(p);
-            case "overlay_show": return _driver.OverlayShow(p);
-            case "overlay_hide": return _driver.OverlayHide(p);
-            case "configure": return _driver.Configure(p);
-            case "approve_app": return _driver.ApproveApp(p);
-            case "end_turn": return _driver.EndTurn(p);
-            case "invalidate_state": return _driver.InvalidateState(p);
-            case "start_audio_recording": return _driver.StartAudioRecording(p);
-            case "stop_audio_recording": return _driver.StopAudioRecording(p);
-            case "shutdown": return new Dictionary<string, object> { { "bye", true } };
-            default: throw new ToolError("unknown_method", "unknown method '" + method + "'");
-        }
-    }
-
-    static string Ok(long id, object result)
-    {
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["id"] = id;
-        d["ok"] = true;
-        d["result"] = result;
-        return Json.ToJson(d);
-    }
-
-    static string Fail(long id, string code, string message)
-    {
-        Dictionary<string, object> err = new Dictionary<string, object>();
-        err["code"] = code;
-        err["message"] = message;
-        Dictionary<string, object> d = new Dictionary<string, object>();
-        d["id"] = id;
-        d["ok"] = false;
-        d["error"] = err;
-        return Json.ToJson(d);
-    }
-}
+        try { if (_driver != null) _driver.Dispose(); } catch { }
+        return RequestedExitCode;
+    }
+
+    static IPlatformDriver CreateDriver()
+    {
+        string forced = Environment.GetEnvironmentVariable("DSH_CU_PLATFORM");
+        if (!string.IsNullOrEmpty(forced) && forced != "windows")
+            throw new NotSupportedException("platform '" + forced + "' has no driver implementation yet; only windows is available");
+        return new WindowsDriver();
+    }
+
+    static int SelfTest()
+    {
+        Dictionary<string, object> ping = _driver.Ping(new Dictionary<string, object>());
+        Dictionary<string, object> wins = _driver.ListWindows(new Dictionary<string, object>());
+        Dictionary<string, object> report = new Dictionary<string, object>();
+        report["ok"] = true;
+        report["platform"] = _driver.PlatformName;
+        report["windowCount"] = wins["count"];
+        report["driver"] = ping;
+        Console.Out.Write(Json.ToJson(report));
+        Console.Out.Write('\n');
+        Console.Out.Flush();
+        return 0;
+    }
+
+    static string Handle(string line)
+    {
+        long id = 0;
+        try
+        {
+            Dictionary<string, object> req = Json.AsObject(Json.Parse(line));
+            if (req == null) throw new ToolError("bad_request", "request must be a JSON object");
+            id = Json.GetLong(req, "id", 0);
+            string method = Json.GetString(req, "method");
+            Dictionary<string, object> p = Json.AsObject(Json.Get(req, "params"));
+            if (p == null) p = new Dictionary<string, object>();
+            // Codex ships the budget with every call (x-oai-cua-request-budget-ms) and the
+            // helper honours it; here it bounds capture waits and marks overruns.
+            Budget.Begin(Json.GetLong(p, "__budgetMs", 0), Json.GetString(p, "__cancelFile"));
+            try { if (_driver != null) _driver.Maintenance(); } catch { }
+            object result = Dispatch(method, p);
+            return Ok(id, result);
+        }
+        catch (ToolError te)
+        {
+            try { _driver.InvalidateState(new Dictionary<string, object>()); } catch { }
+            return Fail(id, te.Code, te.Message);
+        }
+        catch (Exception ex)
+        {
+            try { _driver.InvalidateState(new Dictionary<string, object>()); } catch { }
+            return Fail(id, "internal_error", ex.GetType().Name + ": " + ex.Message);
+        }
+    }
+
+    static object Dispatch(string method, Dictionary<string, object> p)
+    {
+        if (string.IsNullOrEmpty(method)) throw new ToolError("bad_request", "request.method is required");
+        switch (method)
+        {
+            case "ping": return _driver.Ping(p);
+            case "hello": return _driver.Hello(p);
+            case "restore_cursor": return _driver.RestoreCursor(p);
+            case "close": return _driver.Close(p);
+            case "list_windows": return _driver.ListWindows(p);
+            case "list_apps": return _driver.ListApps(p);
+            case "get_window": return _driver.GetWindow(p);
+            case "launch_app": return _driver.LaunchApp(p);
+            case "get_window_state": return _driver.GetWindowState(p);
+            case "click": return _driver.Click(p);
+            case "press_key": return _driver.PressKey(p);
+            case "type_text": return _driver.TypeText(p);
+            case "scroll": return _driver.Scroll(p);
+            case "drag": return _driver.Drag(p);
+            case "set_value": return _driver.SetValue(p);
+            case "perform_secondary_action": return _driver.PerformSecondaryAction(p);
+            case "activate_window": return _driver.ActivateWindow(p);
+            case "overlay_show": return _driver.OverlayShow(p);
+            case "overlay_hide": return _driver.OverlayHide(p);
+            case "configure": return _driver.Configure(p);
+            case "approve_app": return _driver.ApproveApp(p);
+            case "end_turn": return _driver.EndTurn(p);
+            case "invalidate_state": return _driver.InvalidateState(p);
+            case "start_audio_recording": return _driver.StartAudioRecording(p);
+            case "stop_audio_recording": return _driver.StopAudioRecording(p);
+            case "shutdown": return new Dictionary<string, object> { { "bye", true } };
+            default: throw new ToolError("unknown_method", "unknown method '" + method + "'");
+        }
+    }
+
+    static string Ok(long id, object result)
+    {
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["id"] = id;
+        d["ok"] = true;
+        d["result"] = result;
+        return Json.ToJson(d);
+    }
+
+    static string Fail(long id, string code, string message)
+    {
+        Dictionary<string, object> err = new Dictionary<string, object>();
+        err["code"] = code;
+        err["message"] = message;
+        Dictionary<string, object> d = new Dictionary<string, object>();
+        d["id"] = id;
+        d["ok"] = false;
+        d["error"] = err;
+        return Json.ToJson(d);
+    }
+}

package/native/audio/Program.cs

@@ -1,95 +1,95 @@
-using System;
-using System.IO;
-using System.Threading;
-using NAudio.Wave;
-using NAudio.Wave.SampleProviders;
-
-namespace DshCuAudio
-{
-    /// WASAPI loopback recorder: captures whatever the machine is playing, which is
-    /// what Codex's start/stop_audio_recording expose (gated behind SKY_ENABLE_AUDIO).
-    /// Stops on max duration, on the helper's stop file, or on Ctrl+C.
-    internal static class Program
-    {
-        private static int Main(string[] args)
-        {
-            string output = Path.Combine(Path.GetTempPath(), "dsh-cu-audio.wav");
-            int maxMs = 60000;
-            string stopFile = null;
-            for (int i = 0; i < args.Length - 1; i++)
-            {
-                if (args[i] == "--out") output = args[i + 1];
-                else if (args[i] == "--max-ms") int.TryParse(args[i + 1], out maxMs);
-                else if (args[i] == "--stop-file") stopFile = args[i + 1];
-            }
-            if (maxMs < 500) maxMs = 500;
-
-            try
-            {
-                // Record at the device mix format first, then normalise to the 24 kHz
-                // stereo 16-bit WAV that Codex returns.
-                string devicePath = output + ".device.wav";
-                try { if (File.Exists(devicePath)) File.Delete(devicePath); } catch { }
-                using (var capture = new WasapiLoopbackCapture())
-                using (var writer = new WaveFileWriter(devicePath, capture.WaveFormat))
-                {
-                    var stopped = new ManualResetEventSlim(false);
-                    Exception failure = null;
-                    capture.DataAvailable += (sender, e) =>
-                    {
-                        try
-                        {
-                            if (e.BytesRecorded > 0) writer.Write(e.Buffer, 0, e.BytesRecorded);
-                        }
-                        catch (Exception ex) { failure = ex; stopped.Set(); }
-                    };
-                    capture.RecordingStopped += (sender, e) => { stopped.Set(); };
-
-                    capture.StartRecording();
-                    Console.WriteLine("recording started: " + capture.WaveFormat.SampleRate + "Hz " +
-                        capture.WaveFormat.Channels + "ch -> " + output);
-
-                    var deadline = DateTime.UtcNow.AddMilliseconds(maxMs);
-                    while (DateTime.UtcNow < deadline && !stopped.IsSet)
-                    {
-                        if (stopFile != null && File.Exists(stopFile)) break;
-                        Thread.Sleep(150);
-                    }
-
-                    capture.StopRecording();
-                    stopped.Wait(3000);
-                    writer.Flush();
-
-                    if (failure != null) { Console.Error.WriteLine(failure.GetType().Name + ": " + failure.Message); return 6; }
-
-                    writer.Dispose();
-                    Normalize(devicePath, output);
-                    try { File.Delete(devicePath); } catch { }
-                    long bytes = new FileInfo(output).Length;
-                    Console.WriteLine("recorded 24kHz stereo, " + bytes + " bytes -> " + output);
-                    return 0;
-                }
-            }
-            catch (Exception ex)
-            {
-                Console.Error.WriteLine(ex.GetType().Name + ": " + ex.Message);
-                return 5;
-            }
-        }
-
-        /// Resample/convert the device-format capture into 24 kHz stereo 16-bit PCM.
-        static void Normalize(string source, string destination)
-        {
-            using (var reader = new WaveFileReader(source))
-            {
-                ISampleProvider samples = reader.ToSampleProvider();
-                if (samples.WaveFormat.Channels == 1)
-                    samples = new MonoToStereoSampleProvider(samples);
-                ISampleProvider resampled = samples.WaveFormat.SampleRate == 24000
-                    ? samples
-                    : new WdlResamplingSampleProvider(samples, 24000);
-                WaveFileWriter.CreateWaveFile16(destination, resampled);
-            }
-        }
-    }
-}
+using System;
+using System.IO;
+using System.Threading;
+using NAudio.Wave;
+using NAudio.Wave.SampleProviders;
+
+namespace DshCuAudio
+{
+    /// WASAPI loopback recorder: captures whatever the machine is playing, which is
+    /// what Codex's start/stop_audio_recording expose (gated behind SKY_ENABLE_AUDIO).
+    /// Stops on max duration, on the helper's stop file, or on Ctrl+C.
+    internal static class Program
+    {
+        private static int Main(string[] args)
+        {
+            string output = Path.Combine(Path.GetTempPath(), "dsh-cu-audio.wav");
+            int maxMs = 60000;
+            string stopFile = null;
+            for (int i = 0; i < args.Length - 1; i++)
+            {
+                if (args[i] == "--out") output = args[i + 1];
+                else if (args[i] == "--max-ms") int.TryParse(args[i + 1], out maxMs);
+                else if (args[i] == "--stop-file") stopFile = args[i + 1];
+            }
+            if (maxMs < 500) maxMs = 500;
+
+            try
+            {
+                // Record at the device mix format first, then normalise to the 24 kHz
+                // stereo 16-bit WAV that Codex returns.
+                string devicePath = output + ".device.wav";
+                try { if (File.Exists(devicePath)) File.Delete(devicePath); } catch { }
+                using (var capture = new WasapiLoopbackCapture())
+                using (var writer = new WaveFileWriter(devicePath, capture.WaveFormat))
+                {
+                    var stopped = new ManualResetEventSlim(false);
+                    Exception failure = null;
+                    capture.DataAvailable += (sender, e) =>
+                    {
+                        try
+                        {
+                            if (e.BytesRecorded > 0) writer.Write(e.Buffer, 0, e.BytesRecorded);
+                        }
+                        catch (Exception ex) { failure = ex; stopped.Set(); }
+                    };
+                    capture.RecordingStopped += (sender, e) => { stopped.Set(); };
+
+                    capture.StartRecording();
+                    Console.WriteLine("recording started: " + capture.WaveFormat.SampleRate + "Hz " +
+                        capture.WaveFormat.Channels + "ch -> " + output);
+
+                    var deadline = DateTime.UtcNow.AddMilliseconds(maxMs);
+                    while (DateTime.UtcNow < deadline && !stopped.IsSet)
+                    {
+                        if (stopFile != null && File.Exists(stopFile)) break;
+                        Thread.Sleep(150);
+                    }
+
+                    capture.StopRecording();
+                    stopped.Wait(3000);
+                    writer.Flush();
+
+                    if (failure != null) { Console.Error.WriteLine(failure.GetType().Name + ": " + failure.Message); return 6; }
+
+                    writer.Dispose();
+                    Normalize(devicePath, output);
+                    try { File.Delete(devicePath); } catch { }
+                    long bytes = new FileInfo(output).Length;
+                    Console.WriteLine("recorded 24kHz stereo, " + bytes + " bytes -> " + output);
+                    return 0;
+                }
+            }
+            catch (Exception ex)
+            {
+                Console.Error.WriteLine(ex.GetType().Name + ": " + ex.Message);
+                return 5;
+            }
+        }
+
+        /// Resample/convert the device-format capture into 24 kHz stereo 16-bit PCM.
+        static void Normalize(string source, string destination)
+        {
+            using (var reader = new WaveFileReader(source))
+            {
+                ISampleProvider samples = reader.ToSampleProvider();
+                if (samples.WaveFormat.Channels == 1)
+                    samples = new MonoToStereoSampleProvider(samples);
+                ISampleProvider resampled = samples.WaveFormat.SampleRate == 24000
+                    ? samples
+                    : new WdlResamplingSampleProvider(samples, 24000);
+                WaveFileWriter.CreateWaveFile16(destination, resampled);
+            }
+        }
+    }
+}