meeting-assistant/MeetingAssistant/Transcription/AsrDiagnosticService.cs

using MeetingAssistant.Recording;
using NAudio.Wave;

namespace MeetingAssistant.Transcription;

public sealed class AsrDiagnosticService
{
    private readonly ISpeechRecognitionPipelineFactory pipelineFactory;

    public AsrDiagnosticService(ISpeechRecognitionPipelineFactory pipelineFactory)
    {
        this.pipelineFactory = pipelineFactory;
    }

    public Task<AsrDiagnosticResult> TranscribeWavAsync(string path, CancellationToken cancellationToken)
    {
        return TranscribeWavAsync(path, launchProfileName: null, cancellationToken);
    }

    public async Task<AsrDiagnosticResult> TranscribeWavAsync(
        string path,
        string? launchProfileName,
        CancellationToken cancellationToken)
    {
        var fullPath = ResolveExistingWavPath(path);
        await using var pipeline = pipelineFactory.Create(launchProfileName);
        await pipeline.InitializeAsync(cancellationToken);
        var liveTranscriptTask = CollectLiveTranscriptAsync(pipeline, cancellationToken);
        await WriteWavToPipelineAsync(fullPath, pipeline, paceAudio: true, cancellationToken);
        await pipeline.CompleteAsync(cancellationToken);
        return new AsrDiagnosticResult(fullPath, await liveTranscriptTask);
    }

    public async Task<AsrDiagnosticResult> DiarizeWavAsync(
        string path,
        SpeechRecognitionPipelineOptions options,
        CancellationToken cancellationToken)
    {
        var fullPath = ResolveExistingWavPath(path);
        return await DiarizeWavAsync(fullPath, options, launchProfileName: null, cancellationToken);
    }

    public async Task<AsrDiagnosticResult> DiarizeWavAsync(
        string path,
        SpeechRecognitionPipelineOptions options,
        string? launchProfileName,
        CancellationToken cancellationToken)
    {
        var fullPath = ResolveExistingWavPath(path);
        await using var pipeline = pipelineFactory.Create(launchProfileName);
        await pipeline.InitializeAsync(cancellationToken);
        var liveTranscriptTask = CollectLiveTranscriptAsync(pipeline, cancellationToken);
        await WriteWavToPipelineAsync(fullPath, pipeline, paceAudio: true, cancellationToken);
        await pipeline.CompleteAsync(cancellationToken);
        await liveTranscriptTask;
        var finishedSegments = await pipeline.ReadFinishedTranscriptAsync(fullPath, options, cancellationToken);

        return new AsrDiagnosticResult(fullPath, finishedSegments);
    }

    private static async Task<IReadOnlyList<TranscriptionSegment>> CollectLiveTranscriptAsync(
        ISpeechRecognitionPipeline pipeline,
        CancellationToken cancellationToken)
    {
        var segments = new List<TranscriptionSegment>();
        await foreach (var segment in pipeline.ReadLiveTranscriptAsync(cancellationToken))
        {
            segments.Add(segment);
        }

        return segments;
    }

    private static async Task WriteWavToPipelineAsync(
        string fullPath,
        ISpeechRecognitionPipeline pipeline,
        bool paceAudio,
        CancellationToken cancellationToken)
    {
        await foreach (var chunk in ReadPcmWavChunks(fullPath, cancellationToken))
        {
            await pipeline.WriteAsync(chunk, cancellationToken);
            if (paceAudio && chunk.Duration > TimeSpan.Zero)
            {
                await Task.Delay(chunk.Duration, cancellationToken);
            }
        }
    }

    private static string ResolveExistingWavPath(string path)
    {
        if (string.IsNullOrWhiteSpace(path))
        {
            throw new ArgumentException("A WAV file path is required.", nameof(path));
        }

        var fullPath = Path.GetFullPath(Environment.ExpandEnvironmentVariables(path));
        if (!File.Exists(fullPath))
        {
            throw new FileNotFoundException($"WAV file was not found at '{fullPath}'.", fullPath);
        }

        return fullPath;
    }

    private static async IAsyncEnumerable<AudioChunk> ReadPcmWavChunks(
        string path,
        [System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken = default)
    {
        var reader = new WaveFileReader(path);
        try
        {
        var format = reader.WaveFormat;
        if (format.Encoding != WaveFormatEncoding.Pcm || format.BitsPerSample != 16)
        {
            throw new InvalidDataException(
                $"Only 16-bit PCM WAV files are supported. Found {format.Encoding} with {format.BitsPerSample} bits per sample.");
        }

        var buffer = new byte[format.AverageBytesPerSecond];
        int read;
        while ((read = await reader.ReadAsync(buffer.AsMemory(0, buffer.Length), cancellationToken)) > 0)
        {
            yield return new AudioChunk(buffer[..read], format.SampleRate, format.Channels);
        }
        }
        finally
        {
            try
            {
                reader.Dispose();
            }
            catch (NotSupportedException)
            {
                // NAudio can throw while disposing reader streams from async iterators; diagnostics should not fail after reading all chunks.
            }
        }
    }
}

public sealed record AsrDiagnosticResult(string Path, IReadOnlyList<TranscriptionSegment> Segments);