meeting-assistant/MeetingAssistant.Tests/AsrDiagnosticEndpointTests.cs

using System.Net;
using System.Net.Http.Json;
using MeetingAssistant.Recording;
using MeetingAssistant.Speakers;
using MeetingAssistant.Transcription;
using Microsoft.AspNetCore.Mvc.Testing;
using Microsoft.AspNetCore.TestHost;
using Microsoft.Extensions.Configuration;
using Microsoft.Extensions.DependencyInjection;
using Microsoft.Extensions.DependencyInjection.Extensions;

namespace MeetingAssistant.Tests;

public sealed class AsrDiagnosticEndpointTests
{
    [Fact]
    public async Task EndpointTranscribesLocalWavThroughConfiguredProvider()
    {
        await using var factory = CreateFactory<EndpointFakeTranscriptionProvider>();
        using var client = factory.CreateClient();
        var wavPath = Path.Combine(AppContext.BaseDirectory, "Fixtures", "sample-16khz-mono.wav");

        using var response = await client.PostAsJsonAsync("/asr/transcribe-file", new { path = wavPath });
        var body = await response.Content.ReadFromJsonAsync<AsrDiagnosticResponse>();

        Assert.Equal(HttpStatusCode.OK, response.StatusCode);
        Assert.Equal(wavPath, body?.Path);
        var segment = Assert.Single(body?.Segments ?? []);
        Assert.Equal("Unknown", segment.Speaker);
        Assert.StartsWith("endpoint-bytes:", segment.Text, StringComparison.Ordinal);
    }

    [Fact]
    public async Task NamedProfileEndpointPassesProfileToConfiguredProvider()
    {
        var pipelineFactory = new CapturingProfileSpeechRecognitionPipelineFactory();
        await using var factory = new WebApplicationFactory<Program>()
            .WithWebHostBuilder(builder =>
            {
                builder.ConfigureAppConfiguration((_, configuration) =>
                {
                    configuration.AddInMemoryCollection(new Dictionary<string, string?>
                    {
                        ["MeetingAssistant:FunAsr:Backend:Enabled"] = "false",
                        ["MeetingAssistant:LaunchProfiles:english:Hotkey:Toggle"] = "Ctrl+Alt+L",
                        ["MeetingAssistant:LaunchProfiles:english:AzureSpeech:Language"] = "en-US"
                    });
                });
                builder.ConfigureTestServices(services =>
                {
                    services.RemoveAll<ISpeechRecognitionPipelineFactory>();
                    services.AddSingleton<ISpeechRecognitionPipelineFactory>(pipelineFactory);
                });
            });
        using var client = factory.CreateClient();
        var wavPath = Path.Combine(AppContext.BaseDirectory, "Fixtures", "sample-16khz-mono.wav");

        using var response = await client.PostAsJsonAsync("/profiles/english/asr/transcribe-file", new { path = wavPath });

        Assert.Equal(HttpStatusCode.OK, response.StatusCode);
        Assert.Equal("english", pipelineFactory.LastProfileName);
    }

    [Fact]
    public async Task EndpointReportsProviderFailuresAsBadGateway()
    {
        await using var factory = CreateFactory<FailingTranscriptionProvider>();
        using var client = factory.CreateClient();
        var wavPath = Path.Combine(AppContext.BaseDirectory, "Fixtures", "sample-16khz-mono.wav");

        using var response = await client.PostAsJsonAsync("/asr/transcribe-file", new { path = wavPath });

        Assert.Equal(HttpStatusCode.BadGateway, response.StatusCode);
    }

    [Fact]
    public async Task EndpointRunsFinalDiarizationForLocalWav()
    {
        SpeechRecognitionPipelineOptions? capturedOptions = null;
        await using var factory = CreateFactory<EndpointFakeTranscriptionProvider>(
            (audioPath, liveSegments, options, cancellationToken) =>
            {
                capturedOptions = options;
                return Task.FromResult<IReadOnlyList<TranscriptionSegment>>(
                    liveSegments
                        .Select(segment => segment with { Speaker = "Speaker 0", Text = $"diarized:{segment.Text}" })
                        .ToList());
            });
        using var client = factory.CreateClient();
        var wavPath = Path.Combine(AppContext.BaseDirectory, "Fixtures", "sample-16khz-mono.wav");

        using var response = await client.PostAsJsonAsync("/asr/diarize-file", new { path = wavPath, numSpeakers = 5 });
        var body = await response.Content.ReadFromJsonAsync<AsrDiagnosticResponse>();

        Assert.Equal(HttpStatusCode.OK, response.StatusCode);
        Assert.Equal(5, capturedOptions?.NumSpeakers);
        Assert.Equal(wavPath, body?.Path);
        var segment = Assert.Single(body?.Segments ?? []);
        Assert.Equal("Speaker 0", segment.Speaker);
        Assert.StartsWith("diarized:endpoint-bytes:", segment.Text, StringComparison.Ordinal);
    }

    [Fact]
    public async Task SpeakerIdentityMergeDiagnosticEndpointUsesRecentDaysParameter()
    {
        var mergeService = new CapturingSpeakerIdentityMergeService();
        await using var factory = new WebApplicationFactory<Program>()
            .WithWebHostBuilder(builder =>
            {
                builder.ConfigureAppConfiguration((_, configuration) =>
                {
                    configuration.AddInMemoryCollection(new Dictionary<string, string?>
                    {
                        ["MeetingAssistant:FunAsr:Backend:Enabled"] = "false"
                    });
                });
                builder.ConfigureTestServices(services =>
                {
                    services.RemoveAll<ISpeakerIdentityMergeService>();
                    services.AddSingleton<ISpeakerIdentityMergeService>(mergeService);
                });
            });
        using var client = factory.CreateClient();

        using var response = await client.PostAsJsonAsync(
            "/diagnostics/speaker-identities/merge",
            new { recentDays = 7 });

        Assert.Equal(HttpStatusCode.OK, response.StatusCode);
        Assert.Equal(TimeSpan.FromDays(7), mergeService.RecentIdentityAge);
    }

    private static WebApplicationFactory<Program> CreateFactory<TProvider>()
        where TProvider : class, IStreamingTranscriptionProvider
    {
        return CreateFactory<TProvider>((_, _, _, _) => Task.FromResult<IReadOnlyList<TranscriptionSegment>>([]));
    }

    private static WebApplicationFactory<Program> CreateFactory<TProvider>(
        Func<string, IReadOnlyList<TranscriptionSegment>, SpeechRecognitionPipelineOptions, CancellationToken, Task<IReadOnlyList<TranscriptionSegment>>> finalize)
        where TProvider : class, IStreamingTranscriptionProvider
    {
        return new WebApplicationFactory<Program>()
            .WithWebHostBuilder(builder =>
            {
                builder.ConfigureAppConfiguration((_, configuration) =>
                {
                    configuration.AddInMemoryCollection(new Dictionary<string, string?>
                    {
                        ["MeetingAssistant:FunAsr:Backend:Enabled"] = "false"
                    });
                });
                builder.ConfigureTestServices(services =>
                {
                    services.RemoveAll<ISpeechRecognitionPipelineFactory>();
                    services.AddSingleton<ISpeechRecognitionPipelineFactory>(
                        new TestSpeechRecognitionPipelineFactory<TProvider>(finalize));
                });
            });
    }

    private sealed class EndpointFakeTranscriptionProvider : IStreamingTranscriptionProvider
    {
        public async IAsyncEnumerable<TranscriptionSegment> TranscribeAsync(
            IAsyncEnumerable<AudioChunk> audio,
            SpeechRecognitionPipelineOptions options,
            [System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken)
        {
            var bytes = 0;
            await foreach (var chunk in audio.WithCancellation(cancellationToken))
            {
                bytes += chunk.Pcm.Length;
            }

            yield return new TranscriptionSegment(TimeSpan.Zero, TimeSpan.Zero, "Unknown", $"endpoint-bytes:{bytes}");
        }
    }

    private sealed class FailingTranscriptionProvider : IStreamingTranscriptionProvider
    {
        public async IAsyncEnumerable<TranscriptionSegment> TranscribeAsync(
            IAsyncEnumerable<AudioChunk> audio,
            SpeechRecognitionPipelineOptions options,
            [System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken)
        {
            await Task.Yield();
            throw new InvalidOperationException("backend unavailable");
#pragma warning disable CS0162
            yield break;
#pragma warning restore CS0162
        }
    }

    private sealed class TestSpeechRecognitionPipelineFactory<TProvider> : ISpeechRecognitionPipelineFactory
        where TProvider : class, IStreamingTranscriptionProvider
    {
        private readonly Func<string, IReadOnlyList<TranscriptionSegment>, SpeechRecognitionPipelineOptions, CancellationToken, Task<IReadOnlyList<TranscriptionSegment>>> finalize;

        public TestSpeechRecognitionPipelineFactory(
            Func<string, IReadOnlyList<TranscriptionSegment>, SpeechRecognitionPipelineOptions, CancellationToken, Task<IReadOnlyList<TranscriptionSegment>>> finalize)
        {
            this.finalize = finalize;
        }

        public ISpeechRecognitionPipeline Create()
        {
            return new TestSpeechRecognitionPipeline(
                Activator.CreateInstance<TProvider>(),
                finalize);
        }
    }

    private sealed class CapturingProfileSpeechRecognitionPipelineFactory : ISpeechRecognitionPipelineFactory
    {
        public string? LastProfileName { get; private set; }

        public ISpeechRecognitionPipeline Create()
        {
            return Create(null);
        }

        public ISpeechRecognitionPipeline Create(string? launchProfileName)
        {
            LastProfileName = launchProfileName;
            return new TestSpeechRecognitionPipeline(
                new EndpointFakeTranscriptionProvider(),
                (_, liveSegments, _, _) => Task.FromResult(liveSegments));
        }
    }

    private sealed class TestSpeechRecognitionPipeline : StreamingSpeechRecognitionPipeline
    {
        private readonly Func<string, IReadOnlyList<TranscriptionSegment>, SpeechRecognitionPipelineOptions, CancellationToken, Task<IReadOnlyList<TranscriptionSegment>>> finalize;

        public TestSpeechRecognitionPipeline(
            IStreamingTranscriptionProvider provider,
            Func<string, IReadOnlyList<TranscriptionSegment>, SpeechRecognitionPipelineOptions, CancellationToken, Task<IReadOnlyList<TranscriptionSegment>>> finalize)
            : base(provider)
        {
            this.finalize = finalize;
        }

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

    private sealed record AsrDiagnosticResponse(string Path, IReadOnlyList<AsrDiagnosticSegment> Segments);

    private sealed record AsrDiagnosticSegment(string Speaker, string Text);

    private sealed class CapturingSpeakerIdentityMergeService : ISpeakerIdentityMergeService
    {
        public TimeSpan? RecentIdentityAge { get; private set; }

        public Task<SpeakerIdentityMergeResult> MergeRecentIdentitiesAsync(
            TimeSpan? recentIdentityAge,
            CancellationToken cancellationToken)
        {
            RecentIdentityAge = recentIdentityAge;
            return Task.FromResult(new SpeakerIdentityMergeResult(0, 0, 0, 0));
        }
    }
}