Add managed macOS Core Audio voice path

clients/apple/dotnet/README.md

@@ -2,14 +2,15 @@
 
 `VoiceCat.Mac` is the native AppKit C# port. It targets `net10.0-macos` and references the same `VoiceCat.Core` and `VoiceCat.Audio` assemblies used by the Windows client and managed CLI.
 
-The current checkpoint is a functional guest-client shell: AppKit launch/menu lifecycle, host and nickname entry, explicit TOFU approval (including changed-key warning), channel selection, user roster, channel text, voice subscription, disconnect reporting, native accessibility labels, sandbox/network/audio entitlements, and deterministic disposal. It does not yet replace the Swift release. CoreAudio capture/playback, saved servers/accounts, moderation sheets, private messages, settings, ScreenCaptureKit sharing, VoiceOver verification, signing and notarization remain.
+The current checkpoint is a functional guest client: AppKit launch/menu lifecycle, host and nickname entry, explicit TOFU approval (including changed-key warning), channel selection, user roster, channel text, voice subscription, managed microphone-stream lifecycle, default-device Core Audio capture/playback, disconnect reporting, native accessibility labels, sandbox/network/audio entitlements, and deterministic disposal. Capture is converted to interleaved 48 kHz int16 PCM through `AVAudioConverter`; playback uses an `AVAudioSourceNode` and the shared bounded `PcmRing`, so its render callback does not allocate, lock, or block. It does not yet replace the Swift release. Saved servers/accounts, non-default device selection, moderation sheets, private messages, settings, ScreenCaptureKit sharing, VoiceOver verification, signing and notarization remain.
 
 Build on Apple Silicon macOS 15.6+ with Xcode 26 and the .NET 10 macOS workload:
 
 ```bash
 dotnet workload install macos
+pwsh ./dotnet/build-native.ps1
 dotnet restore clients/apple/dotnet/VoiceCat.Apple.slnx
 dotnet build clients/apple/dotnet/VoiceCat.Apple.slnx -c Debug
 ```
 
-The repository's `dotnet/build-native.ps1` must first stage an `osx-arm64` `libvoicecat_media.dylib`; that shim contains only Opus/RNNoise. Windows and Linux cannot compile or validate AppKit bindings, so the macOS build is a required CI/release gate.
+The native build stages an `osx-arm64` `libvoicecat_media.dylib`; that shim contains only Opus/RNNoise. The AppKit and AVFoundation calls are also compiled against Microsoft's current macOS reference assembly during development, but only a macOS host can link, launch, grant microphone access and verify live devices. The macOS CI build and manual listen test are required release gates.

clients/apple/dotnet/VoiceCat.Mac/AppDelegate.cs

@@ -11,7 +11,6 @@ internal sealed class AppDelegate : NSApplicationDelegate
         NSApplication.SharedApplication.ActivationPolicy = NSApplicationActivationPolicy.Regular;
         BuildMenu();
         connect = new(); connect.ShowWindow(this);
-        NSApplication.SharedApplication.ActivateIgnoringOtherApps(true);
     }
     public override bool ApplicationShouldTerminateAfterLastWindowClosed(NSApplication sender) => true;
     private static void BuildMenu()

clients/apple/dotnet/VoiceCat.Mac/ConnectWindowController.cs

@@ -9,7 +9,7 @@ namespace VoiceCat.Mac;
 internal sealed class ConnectWindowController : NSWindowController
 {
     private readonly NSTextField host = new(new CGRect(120, 180, 280, 26)) { StringValue = "127.0.0.1:8384", PlaceholderString = "Host:port" };
-    private readonly NSTextField nickname = new(new CGRect(120, 140, 280, 26)) { StringValue = NSProcessInfo.ProcessInfo.UserName, PlaceholderString = "Nickname" };
+    private readonly NSTextField nickname = new(new CGRect(120, 140, 280, 26)) { StringValue = Environment.UserName, PlaceholderString = "Nickname" };
     private readonly NSTextField status = NSTextField.CreateLabel("Ready to connect");
     private readonly NSButton connect = new(new CGRect(300, 55, 100, 32)) { Title = "Connect", BezelStyle = NSBezelStyle.Rounded };
     private VoiceCatClient? client;
@@ -23,7 +23,7 @@ internal sealed class ConnectWindowController : NSWindowController
         var hostLabel = NSTextField.CreateLabel("Server"); hostLabel.Frame = new CGRect(30, 185, 80, 20); view.AddSubview(hostLabel); view.AddSubview(host);
         var nicknameLabel = NSTextField.CreateLabel("Nickname"); nicknameLabel.Frame = new CGRect(30, 145, 80, 20); view.AddSubview(nicknameLabel); view.AddSubview(nickname);
         status.Frame = new CGRect(30, 95, 370, 22); status.AccessibilityLabel = "Connection status"; view.AddSubview(status);
-        connect.AccessibilityLabel = "Connect to server"; connect.Activated += Connect; view.AddSubview(connect);
+        ((INSAccessibility)connect).AccessibilityLabel = "Connect to server"; connect.Activated += Connect; view.AddSubview(connect);
         Window.DefaultButtonCell = connect.Cell;
     }
 

clients/apple/dotnet/VoiceCat.Mac/MacAudioBackend.cs

@@ -0,0 +1,200 @@
+using System.Runtime.InteropServices;
+using AVFoundation;
+using Foundation;
+using VoiceCat.Audio;
+
+namespace VoiceCat.Mac;
+
+internal sealed class MacAudioBackend : IAudioDeviceBackend
+{
+    private static readonly AudioDeviceInfo[] DefaultInput = [new("", "System default microphone", true)];
+    private static readonly AudioDeviceInfo[] DefaultOutput = [new("", "System default output", true)];
+    private Exception? failure;
+
+    internal Exception? Failure => Volatile.Read(ref failure);
+
+    public IReadOnlyList<AudioDeviceInfo> Enumerate(bool input) => input ? DefaultInput : DefaultOutput;
+
+    public IAudioCapture OpenCapture(string? deviceId, bool loopback, CapturePcmHandler pcm)
+    {
+        if (loopback) throw new NotSupportedException("Screen audio uses ScreenCaptureKit, not microphone capture.");
+        if (!string.IsNullOrEmpty(deviceId)) throw new NotSupportedException("Selecting a non-default Core Audio input is not implemented yet.");
+        return new Capture(pcm, ReportFailure);
+    }
+
+    public IAudioPlayback OpenPlayback(string? deviceId = null)
+    {
+        if (!string.IsNullOrEmpty(deviceId)) throw new NotSupportedException("Selecting a non-default Core Audio output is not implemented yet.");
+        return new Playback(ReportFailure);
+    }
+
+    private void ReportFailure(Exception exception) => Interlocked.CompareExchange(ref failure, exception, null);
+
+    private sealed class Capture : IAudioCapture
+    {
+        private readonly CapturePcmHandler handler;
+        private readonly AVAudioEngine engine = new();
+        private readonly AVAudioFormat outputFormat;
+        private readonly AVAudioConverter converter;
+        private readonly AVAudioPcmBuffer converted;
+        private readonly AVAudioConverterInputHandler inputProvider;
+        private readonly Action<Exception> reportFailure;
+        private AVAudioPcmBuffer? pendingInput;
+        private bool inputProvided;
+        private int disposed;
+
+        internal Capture(CapturePcmHandler handler, Action<Exception> reportFailure)
+        {
+            this.handler = handler;
+            this.reportFailure = reportFailure;
+            AVAudioInputNode input = engine.InputNode;
+            AVAudioFormat inputFormat = input.GetBusOutputFormat(0);
+            uint channels = Math.Clamp(inputFormat.ChannelCount, 1u, 2u);
+            outputFormat = new(AVAudioCommonFormat.PCMInt16, 48_000, channels, true);
+            converter = new(inputFormat, outputFormat);
+            uint outputCapacity = checked((uint)Math.Ceiling(4_096 * 48_000 / inputFormat.SampleRate) + 64);
+            converted = new(outputFormat, outputCapacity);
+            inputProvider = ProvideInput;
+            input.InstallTapOnBus(0, 960, inputFormat, Convert);
+            engine.Prepare();
+            if (!engine.StartAndReturnError(out var error))
+            {
+                input.RemoveTapOnBus(0);
+                converted.Dispose();
+                converter.Dispose();
+                outputFormat.Dispose();
+                engine.Dispose();
+                throw new InvalidOperationException("Core Audio capture could not start: " + error.LocalizedDescription);
+            }
+        }
+
+        private unsafe void Convert(AVAudioPcmBuffer input, AVAudioTime _time)
+        {
+            try
+            {
+                pendingInput = input;
+                inputProvided = false;
+                converted.FrameLength = 0;
+                AVAudioConverterOutputStatus result = converter.ConvertToBuffer(converted, out NSError? error, inputProvider);
+                if (result == AVAudioConverterOutputStatus.Error)
+                {
+                    reportFailure(new InvalidOperationException("Core Audio input conversion failed: " + error?.LocalizedDescription));
+                    return;
+                }
+                if (converted.FrameLength == 0) return;
+                nint samples = Marshal.ReadIntPtr(converted.Int16ChannelData);
+                if (samples == 0) return;
+                int channels = checked((int)outputFormat.ChannelCount);
+                int remainingFrames = checked((int)converted.FrameLength);
+                var pcm = new ReadOnlySpan<short>((void*)samples, checked(remainingFrames * channels));
+                while (remainingFrames > 0)
+                {
+                    int frames = Math.Min(960, remainingFrames);
+                    handler(pcm[..(frames * channels)], channels);
+                    pcm = pcm[(frames * channels)..];
+                    remainingFrames -= frames;
+                }
+            }
+            catch (Exception exception) { reportFailure(exception); }
+            finally { pendingInput = null; }
+        }
+
+        private AVAudioBuffer ProvideInput(uint _, out AVAudioConverterInputStatus status)
+        {
+            if (!inputProvided && pendingInput is { } input)
+            {
+                inputProvided = true;
+                status = AVAudioConverterInputStatus.HaveData;
+                return input;
+            }
+            status = AVAudioConverterInputStatus.NoDataNow;
+            return null!;
+        }
+
+        public void Dispose()
+        {
+            if (Interlocked.Exchange(ref disposed, 1) != 0) return;
+            engine.InputNode.RemoveTapOnBus(0);
+            engine.Stop();
+            converted.Dispose();
+            converter.Dispose();
+            outputFormat.Dispose();
+            engine.Dispose();
+        }
+    }
+
+    private sealed class Playback : IAudioPlayback
+    {
+        private readonly PcmRing pcm = new(32_768);
+        private readonly AVAudioEngine engine = new();
+        private readonly AVAudioFormat format;
+        private readonly AVAudioSourceNode source;
+        private readonly Action<Exception> reportFailure;
+        private int callbackFailed;
+        private int disposed;
+
+        internal Playback(Action<Exception> reportFailure)
+        {
+            this.reportFailure = reportFailure;
+            format = new(AVAudioCommonFormat.PCMInt16, 48_000, 2, true);
+            source = new(format, Render);
+            engine.AttachNode(source);
+            engine.Connect(source, engine.MainMixerNode, format);
+            engine.Prepare();
+            if (!engine.StartAndReturnError(out var error))
+            {
+                engine.DetachNode(source);
+                source.Dispose();
+                format.Dispose();
+                engine.Dispose();
+                throw new InvalidOperationException("Core Audio playback could not start: " + error.LocalizedDescription);
+            }
+        }
+
+        public void Write(ReadOnlySpan<short> stereoPcm) => pcm.TryWrite(stereoPcm);
+
+        private unsafe int Render(IntPtr isSilence, IntPtr _, uint frameCount, IntPtr outputData)
+        {
+            try
+            {
+                int bufferCount = Marshal.ReadInt32(outputData);
+                int firstBuffer = IntPtr.Size == 8 ? 8 : 4;
+                if (bufferCount != 1) return Fail("Core Audio returned a non-interleaved playback buffer.");
+                int channels = Marshal.ReadInt32(outputData, firstBuffer);
+                int byteCount = Marshal.ReadInt32(outputData, firstBuffer + 4);
+                nint data = Marshal.ReadIntPtr(outputData, firstBuffer + 8);
+                int requested = checked((int)frameCount * 2);
+                if (channels != 2 || data == 0 || byteCount < requested * sizeof(short)) return Fail("Core Audio returned an invalid stereo playback buffer.");
+                var output = new Span<short>((void*)data, requested);
+                Span<short> discard = stackalloc short[1_920];
+                while (pcm.Count > 11_520) pcm.Read(discard[..Math.Min(discard.Length, pcm.Count - 11_520)]);
+                int read = pcm.Read(output);
+                output[read..].Clear();
+                if (isSilence != IntPtr.Zero) Marshal.WriteByte(isSilence, read == 0 ? (byte)1 : (byte)0);
+                return 0;
+            }
+            catch (Exception exception)
+            {
+                if (Interlocked.Exchange(ref callbackFailed, 1) == 0) reportFailure(exception);
+                return -1;
+            }
+        }
+
+        private int Fail(string message)
+        {
+            if (Interlocked.Exchange(ref callbackFailed, 1) == 0) reportFailure(new InvalidOperationException(message));
+            return -1;
+        }
+
+        public void Dispose()
+        {
+            if (Interlocked.Exchange(ref disposed, 1) != 0) return;
+            engine.Stop();
+            engine.DetachNode(source);
+            source.Dispose();
+            format.Dispose();
+            engine.Dispose();
+        }
+    }
+
+}

clients/apple/dotnet/VoiceCat.Mac/MainWindowController.cs

@@ -1,6 +1,7 @@
 using AppKit;
 using CoreGraphics;
 using Foundation;
+using VoiceCat.Audio;
 using VoiceCat.Core;
 using Voicecat.V1;
 
@@ -9,6 +10,7 @@ namespace VoiceCat.Mac;
 internal sealed class MainWindowController : NSWindowController
 {
     private readonly VoiceCatClient client;
+    private readonly MacAudioBackend audioBackend = new();
     private readonly uint selfId;
     private readonly NSPopUpButton channels = new(new CGRect(20, 515, 300, 28), false);
     private readonly NSTextView users = new(new CGRect(0, 0, 190, 430)) { Editable = false, Selectable = true };
@@ -19,7 +21,10 @@ internal sealed class MainWindowController : NSWindowController
     private readonly NSTextField status = NSTextField.CreateLabel("Connected");
     private readonly NSTimer timer;
     private uint currentChannel = 1;
+    private uint microphoneStreamId;
     private bool joinedVoice;
+    private IAudioCapture? microphone;
+    private IAudioPlayback? playback;
 
     internal MainWindowController(VoiceCatClient client, uint selfId, string nickname) : base(new NSWindow(new CGRect(0, 0, 760, 570),
         NSWindowStyle.Titled | NSWindowStyle.Closable | NSWindowStyle.Resizable | NSWindowStyle.Miniaturizable, NSBackingStore.Buffered, false))
@@ -27,12 +32,12 @@ internal sealed class MainWindowController : NSWindowController
         this.client = client; this.selfId = selfId;
         Window!.Title = $"VoiceCat — {nickname}"; Window.Center(); Window.MinSize = new CGSize(680, 480);
         var content = Window.ContentView!;
-        channels.AccessibilityLabel = "Channel"; channels.Activated += ChangeChannel; content.AddSubview(channels);
-        voice.AccessibilityLabel = "Join or leave voice"; voice.Activated += ToggleVoice; content.AddSubview(voice);
+        ((INSAccessibility)channels).AccessibilityLabel = "Channel"; channels.Activated += ChangeChannel; content.AddSubview(channels);
+        ((INSAccessibility)voice).AccessibilityLabel = "Join or leave voice"; voice.Activated += ToggleVoice; content.AddSubview(voice);
         var userScroll = new NSScrollView(new CGRect(20, 90, 190, 410)) { HasVerticalScroller = true, DocumentView = users }; userScroll.AccessibilityLabel = "Users in channel"; content.AddSubview(userScroll);
         var chatScroll = new NSScrollView(new CGRect(230, 90, 500, 410)) { HasVerticalScroller = true, DocumentView = chat }; chatScroll.AccessibilityLabel = "Channel messages"; content.AddSubview(chatScroll);
-        compose.AccessibilityLabel = "Message"; compose.Activated += Send; content.AddSubview(compose);
-        send.AccessibilityLabel = "Send message"; send.Activated += Send; content.AddSubview(send);
+        ((INSAccessibility)compose).AccessibilityLabel = "Message"; compose.Activated += Send; content.AddSubview(compose);
+        ((INSAccessibility)send).AccessibilityLabel = "Send message"; send.Activated += Send; content.AddSubview(send);
         status.Frame = new CGRect(20, 22, 700, 22); status.AccessibilityLabel = "Connection status"; content.AddSubview(status);
         timer = NSTimer.CreateRepeatingScheduledTimer(TimeSpan.FromMilliseconds(50), _ => Pump());
         RefreshState();
@@ -44,13 +49,17 @@ internal sealed class MainWindowController : NSWindowController
         {
             if (envelope!.TextMessage is { } text) Append($"[{DateTime.Now:t}] {Name(text.SenderId)}: {text.Body}");
             if (envelope.ServerState is not null || envelope.ChannelEvent is not null || envelope.UserEvent is not null) RefreshState();
-            if (envelope.Disconnect is { } disconnected) { status.StringValue = "Disconnected: " + disconnected.Reason; voice.Enabled = send.Enabled = false; }
+            if (envelope.Disconnect is { } disconnected)
+            {
+                StopAudioDevices(); joinedVoice = false; voice.Title = "Join voice";
+                status.StringValue = "Disconnected: " + disconnected.Reason; voice.Enabled = send.Enabled = false;
+            }
         }
         if (client.Audio.Failure is { } failure) status.StringValue = "Audio stopped: " + failure.Message;
+        if (audioBackend.Failure is { } deviceFailure) status.StringValue = "Audio device stopped: " + deviceFailure.Message;
     }
     private void RefreshState()
     {
-        string? selected = channels.SelectedItem?.RepresentedObject?.ToString();
         channels.RemoveAllItems();
         foreach (var channel in client.Channels.OrderBy(c => c.Order).ThenBy(c => c.Name)) { channels.AddItem(channel.Name); channels.LastItem!.RepresentedObject = new NSString(channel.Id.ToString()); }
         currentChannel = client.Users.FirstOrDefault(u => u.Id == selfId)?.ChannelId ?? currentChannel;
@@ -61,7 +70,14 @@ internal sealed class MainWindowController : NSWindowController
     private async void ChangeChannel(object? sender, EventArgs args)
     {
         if (!uint.TryParse(channels.SelectedItem?.RepresentedObject?.ToString(), out uint id) || id == currentChannel) return;
-        try { var result = (await client.RequestAsync(new() { JoinChannel = new() { ChannelId = id } })).JoinChannelResult; if (!result.Ok) status.StringValue = result.Error; }
+        try
+        {
+            bool resumeVoice = joinedVoice;
+            if (resumeVoice) await LeaveVoice();
+            var result = (await client.RequestAsync(new() { JoinChannel = new() { ChannelId = id } })).JoinChannelResult;
+            if (!result.Ok) status.StringValue = result.Error;
+            if (resumeVoice) await JoinVoice();
+        }
         catch (Exception exception) { status.StringValue = exception.Message; }
     }
     private void Send(object? sender, EventArgs args)
@@ -74,16 +90,70 @@ internal sealed class MainWindowController : NSWindowController
     {
         try
         {
-            if (!joinedVoice) { var result = await client.SubscribeVoiceAsync(); if (!result.Ok) throw new InvalidOperationException(result.Error); joinedVoice = true; voice.Title = "Leave voice"; }
-            else { await client.RequestAsync(new() { UnsubscribeVoice = new() }); joinedVoice = false; voice.Title = "Join voice"; }
+            voice.Enabled = false;
+            if (!joinedVoice) await JoinVoice();
+            else await LeaveVoice();
         }
         catch (Exception exception) { status.StringValue = exception.Message; }
+        finally { voice.Enabled = true; }
+    }
+
+    private async Task JoinVoice()
+    {
+        var result = await client.SubscribeVoiceAsync();
+        if (!result.Ok) throw new InvalidOperationException(result.Error);
+        try
+        {
+            playback = audioBackend.OpenPlayback();
+            client.Audio.MixedPcm += PlayMixedPcm;
+            StreamInfo stream = await client.StartStreamAsync(StreamKind.StreamMic, "Microphone");
+            microphoneStreamId = stream.StreamId;
+            microphone = audioBackend.OpenCapture(null, false, FeedMicrophone);
+            joinedVoice = true; voice.Title = "Leave voice"; status.StringValue = "Voice connected";
+        }
+        catch
+        {
+            try { if (microphoneStreamId != 0) client.StopStream(microphoneStreamId); } catch { }
+            StopAudioDevices();
+            try { await client.SubscribeVoiceAsync(false); } catch { }
+            throw;
+        }
+    }
+
+    private async Task LeaveVoice()
+    {
+        uint streamId = microphoneStreamId;
+        joinedVoice = false; voice.Title = "Join voice";
+        StopAudioDevices();
+        if (streamId != 0) client.StopStream(streamId);
+        await client.SubscribeVoiceAsync(false);
+        status.StringValue = "Voice disconnected";
+    }
+
+    private void FeedMicrophone(ReadOnlySpan<short> pcm, int channels)
+    {
+        uint streamId = microphoneStreamId;
+        if (streamId != 0) client.Audio.FeedPcm(streamId, pcm, channels);
+    }
+
+    private void PlayMixedPcm(ReadOnlySpan<short> pcm) => Volatile.Read(ref playback)?.Write(pcm);
+
+    private void StopAudioDevices()
+    {
+        microphoneStreamId = 0;
+        Interlocked.Exchange(ref microphone, null)?.Dispose();
+        client.Audio.MixedPcm -= PlayMixedPcm;
+        Interlocked.Exchange(ref playback, null)?.Dispose();
     }
     private string Name(uint id) => client.Users.FirstOrDefault(u => u.Id == id)?.Nickname ?? $"User {id}";
-    private void Append(string line) { chat.Value = chat.Value.Length == 0 ? line : chat.Value + "\n" + line; chat.ScrollToEndOfDocument(this); }
+    private void Append(string line)
+    {
+        chat.Value = chat.Value.Length == 0 ? line : chat.Value + "\n" + line;
+        chat.ScrollRangeToVisible(new NSRange(chat.Value.Length, 0));
+    }
     protected override void Dispose(bool disposing)
     {
-        if (disposing) { timer.Invalidate(); client.DisposeAsync().AsTask().GetAwaiter().GetResult(); }
+        if (disposing) { timer.Invalidate(); StopAudioDevices(); client.DisposeAsync().AsTask().GetAwaiter().GetResult(); }
         base.Dispose(disposing);
     }
 }

clients/apple/dotnet/VoiceCat.Mac/VoiceCat.Mac.csproj

@@ -6,6 +6,7 @@
     <SupportedOSPlatformVersion>14.0</SupportedOSPlatformVersion>
     <Nullable>enable</Nullable>
     <ImplicitUsings>enable</ImplicitUsings>
+    <AllowUnsafeBlocks>true</AllowUnsafeBlocks>
     <ApplicationTitle>VoiceCat</ApplicationTitle>
     <ApplicationId>net.iamtalon.voicecat</ApplicationId>
     <UseHardenedRuntime>true</UseHardenedRuntime>