meeting-assistant/MeetingAssistant.Tests/AsrDiagnosticTests.cs

using MeetingAssistant.Recording;
using MeetingAssistant.Transcription;

namespace MeetingAssistant.Tests;

public sealed class AsrDiagnosticTests
{
    [Fact]
    public async Task DiagnosticStreamsPcmWavThroughConfiguredProvider()
    {
        var provider = new CapturingTranscriptionProvider();
        var diagnostics = new AsrDiagnosticService(
            new TestSpeechRecognitionPipelineFactory(provider));
        var wavPath = Path.Combine(AppContext.BaseDirectory, "Fixtures", "sample-16khz-mono.wav");

        var result = await diagnostics.TranscribeWavAsync(wavPath, CancellationToken.None);

        Assert.Equal(wavPath, result.Path);
        Assert.Equal(16000, provider.Chunks.Single().SampleRate);
        Assert.Equal(1, provider.Chunks.Single().Channels);
        Assert.NotEmpty(provider.Chunks.Single().Pcm);
        var segment = Assert.Single(result.Segments);
        Assert.Equal("Unknown", segment.Speaker);
        Assert.Equal("bytes:" + provider.Chunks.Single().Pcm.Length, segment.Text);
    }

    private sealed class CapturingTranscriptionProvider : IStreamingTranscriptionProvider
    {
        public List<AudioChunk> Chunks { get; } = [];

        public async IAsyncEnumerable<TranscriptionSegment> TranscribeAsync(
            IAsyncEnumerable<AudioChunk> audio,
            SpeechRecognitionPipelineOptions options,
            [System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken)
        {
            await foreach (var chunk in audio.WithCancellation(cancellationToken))
            {
                Chunks.Add(chunk);
            }

            yield return new TranscriptionSegment(
                TimeSpan.Zero,
                TimeSpan.Zero,
                "Unknown",
                "bytes:" + Chunks.Sum(chunk => chunk.Pcm.Length));
        }
    }

    private sealed class TestSpeechRecognitionPipelineFactory : ISpeechRecognitionPipelineFactory
    {
        private readonly IStreamingTranscriptionProvider provider;

        public TestSpeechRecognitionPipelineFactory(IStreamingTranscriptionProvider provider)
        {
            this.provider = provider;
        }

        public ISpeechRecognitionPipeline Create()
        {
            return new TestSpeechRecognitionPipeline(provider);
        }
    }

    private sealed class TestSpeechRecognitionPipeline : StreamingSpeechRecognitionPipeline
    {
        public TestSpeechRecognitionPipeline(IStreamingTranscriptionProvider provider)
            : base(provider)
        {
        }

        protected override Task<IReadOnlyList<TranscriptionSegment>> BuildFinishedTranscriptAsync(
            string audioPath,
            IReadOnlyList<TranscriptionSegment> liveSegments,
            SpeechRecognitionPipelineOptions options,
            CancellationToken cancellationToken)
        {
            return Task.FromResult<IReadOnlyList<TranscriptionSegment>>([]);
        }
    }
}