meeting-assistant/MeetingAssistant/Recording/MacOsMeetingAudioSource.cs

using System.Diagnostics;
using System.Runtime.CompilerServices;

namespace MeetingAssistant.Recording;

internal enum MacOsAudioCaptureKind
{
    Microphone,
    System
}

internal sealed record MacOsAudioCaptureRequest(
    MacOsAudioCaptureKind Kind,
    int SampleRate,
    int Channels);

internal interface IMacOsAudioCaptureProcessFactory
{
    IMacOsAudioCaptureProcess Start(MacOsAudioCaptureRequest request);
}

internal interface IMacOsAudioCaptureProcess : IAsyncDisposable
{
    Stream StandardOutput { get; }

    int ExitCode { get; }

    bool HasExited { get; }

    Task<string> ReadStandardErrorAsync(CancellationToken cancellationToken);

    Task WaitForExitAsync(CancellationToken cancellationToken);

    void Terminate();
}

internal abstract class MacOsMeetingAudioSource : IMeetingAudioSource
{
    private const int ReadBufferSize = 8 * 1024;
    private readonly MacOsAudioCaptureKind kind;
    private readonly IMacOsAudioCaptureProcessFactory processFactory;
    private readonly ILogger logger;

    protected MacOsMeetingAudioSource(
        MacOsAudioCaptureKind kind,
        IMacOsAudioCaptureProcessFactory processFactory,
        ILogger logger)
    {
        this.kind = kind;
        this.processFactory = processFactory;
        this.logger = logger;
    }

    public IAsyncEnumerable<AudioChunk> CaptureAsync(CancellationToken cancellationToken)
    {
        return CaptureAsync(new MeetingAssistantOptions(), cancellationToken);
    }

    public async IAsyncEnumerable<AudioChunk> CaptureAsync(
        MeetingAssistantOptions options,
        [EnumeratorCancellation] CancellationToken cancellationToken)
    {
        var request = new MacOsAudioCaptureRequest(
            kind,
            options.Recording.SampleRate,
            options.Recording.Channels);
        await using var process = processFactory.Start(request);
        var standardErrorTask = process.ReadStandardErrorAsync(CancellationToken.None);
        var buffer = new byte[ReadBufferSize];
        byte? pendingByte = null;
        var hasLoggedFirstChunk = false;

        logger.LogInformation(
            "Starting macOS {CaptureKind} audio capture at {SampleRate} Hz/{Channels} channel(s)",
            kind,
            request.SampleRate,
            request.Channels);

        try
        {
            while (true)
            {
                var bytesRead = await process.StandardOutput.ReadAsync(buffer, cancellationToken);
                if (bytesRead == 0)
                {
                    break;
                }

                var pcm = CopyCompleteSamples(buffer, bytesRead, ref pendingByte);
                if (pcm.Length == 0)
                {
                    continue;
                }

                if (!hasLoggedFirstChunk)
                {
                    hasLoggedFirstChunk = true;
                    logger.LogInformation(
                        "macOS {CaptureKind} audio produced its first PCM chunk: {ByteCount} bytes",
                        kind,
                        pcm.Length);
                }

                yield return new AudioChunk(pcm, request.SampleRate, request.Channels);
            }

            await process.WaitForExitAsync(cancellationToken);
            var standardError = await standardErrorTask;
            if (process.ExitCode != 0)
            {
                throw CreateCaptureException(kind, process.ExitCode, standardError);
            }
        }
        finally
        {
            if (!process.HasExited)
            {
                process.Terminate();
                await process.WaitForExitAsync(CancellationToken.None);
            }
        }
    }

    private static byte[] CopyCompleteSamples(byte[] buffer, int bytesRead, ref byte? pendingByte)
    {
        var totalBytes = bytesRead + (pendingByte.HasValue ? 1 : 0);
        var completeBytes = totalBytes - totalBytes % sizeof(short);
        var pcm = new byte[completeBytes];
        var targetOffset = 0;

        if (pendingByte is { } firstByte)
        {
            pcm[0] = firstByte;
            targetOffset = 1;
            pendingByte = null;
        }

        var copyCount = completeBytes - targetOffset;
        if (copyCount > 0)
        {
            Buffer.BlockCopy(buffer, 0, pcm, targetOffset, copyCount);
        }

        var consumedFromBuffer = copyCount;
        if (consumedFromBuffer < bytesRead)
        {
            pendingByte = buffer[consumedFromBuffer];
        }

        return pcm;
    }

    private static InvalidOperationException CreateCaptureException(
        MacOsAudioCaptureKind kind,
        int exitCode,
        string standardError)
    {
        var detail = string.IsNullOrWhiteSpace(standardError)
            ? "The native capture helper did not provide an error message."
            : standardError.Trim();
        return new InvalidOperationException(
            $"macOS {kind.ToString().ToLowerInvariant()} audio capture exited with code {exitCode}. {detail}");
    }
}

internal sealed class MacOsMicrophoneAudioSource : MacOsMeetingAudioSource
{
    public MacOsMicrophoneAudioSource(
        IMacOsAudioCaptureProcessFactory processFactory,
        ILogger<MacOsMicrophoneAudioSource> logger)
        : base(MacOsAudioCaptureKind.Microphone, processFactory, logger)
    {
    }
}

internal sealed class MacOsSystemAudioSource : MacOsMeetingAudioSource
{
    public MacOsSystemAudioSource(
        IMacOsAudioCaptureProcessFactory processFactory,
        ILogger<MacOsSystemAudioSource> logger)
        : base(MacOsAudioCaptureKind.System, processFactory, logger)
    {
    }
}

internal sealed class MacOsAudioCaptureProcessFactory : IMacOsAudioCaptureProcessFactory
{
    private readonly ILogger<MacOsAudioCaptureProcessFactory> logger;

    public MacOsAudioCaptureProcessFactory(ILogger<MacOsAudioCaptureProcessFactory> logger)
    {
        this.logger = logger;
    }

    public IMacOsAudioCaptureProcess Start(MacOsAudioCaptureRequest request)
    {
        if (!OperatingSystem.IsMacOS())
        {
            throw new PlatformNotSupportedException("The macOS audio capture helper can only run on macOS.");
        }

        var helperPath = Path.Combine(AppContext.BaseDirectory, "Native", "macos-meeting-audio-capture");
        if (!File.Exists(helperPath))
        {
            throw new FileNotFoundException(
                $"The macOS audio capture helper was not found at '{helperPath}'. Build or publish the application on macOS so the Swift helper is compiled into the Native folder.",
                helperPath);
        }

        var startInfo = new ProcessStartInfo
        {
            FileName = helperPath,
            RedirectStandardOutput = true,
            RedirectStandardError = true,
            UseShellExecute = false,
            CreateNoWindow = true
        };
        startInfo.ArgumentList.Add(request.Kind == MacOsAudioCaptureKind.Microphone ? "microphone" : "system");
        startInfo.ArgumentList.Add(request.SampleRate.ToString(System.Globalization.CultureInfo.InvariantCulture));
        startInfo.ArgumentList.Add(request.Channels.ToString(System.Globalization.CultureInfo.InvariantCulture));

        logger.LogDebug("Starting native macOS audio helper {HelperPath} for {CaptureKind}", helperPath, request.Kind);
        var process = Process.Start(startInfo)
            ?? throw new InvalidOperationException($"Failed to start the macOS audio capture helper '{helperPath}'.");
        return new MacOsAudioCaptureProcess(process);
    }

}

internal sealed class MacOsAudioCaptureProcess : IMacOsAudioCaptureProcess
{
    private readonly Process process;

    public MacOsAudioCaptureProcess(Process process)
    {
        this.process = process;
    }

    public Stream StandardOutput => process.StandardOutput.BaseStream;

    public int ExitCode => process.ExitCode;

    public bool HasExited => process.HasExited;

    public Task<string> ReadStandardErrorAsync(CancellationToken cancellationToken)
    {
        return process.StandardError.ReadToEndAsync(cancellationToken);
    }

    public Task WaitForExitAsync(CancellationToken cancellationToken)
    {
        return process.WaitForExitAsync(cancellationToken);
    }

    public void Terminate()
    {
        if (!process.HasExited)
        {
            process.Kill(entireProcessTree: true);
        }
    }

    public ValueTask DisposeAsync()
    {
        process.Dispose();
        return ValueTask.CompletedTask;
    }
}