Add meeting assistant speech and summary automation

This commit is contained in:
2026-05-21 09:55:39 +02:00
parent 1e97d2b9ff
commit 48d98d9cbb
63 changed files with 5143 additions and 151 deletions
@@ -1,4 +1,5 @@
using MeetingAssistant.Recording;
using MeetingAssistant.Speakers;
using MeetingAssistant.Transcription;
using MeetingAssistant.MeetingNotes;
using MeetingAssistant.Summary;
@@ -10,6 +11,22 @@ namespace MeetingAssistant.Tests;
public sealed class RecordingCoordinatorTests
{
private static async Task WaitUntilAsync(Func<bool> condition)
{
var deadline = DateTimeOffset.UtcNow.AddSeconds(5);
while (DateTimeOffset.UtcNow < deadline)
{
if (condition())
{
return;
}
await Task.Delay(25);
}
throw new TimeoutException("Condition was not met.");
}
[Fact]
public async Task ToggleStartsStreamingTranscriptionAndSecondToggleStopsIt()
{
@@ -116,6 +133,7 @@ public sealed class RecordingCoordinatorTests
DateTimeOffset.Parse("2026-05-19T11:00:00+02:00"))));
await coordinator.StartAsync(CancellationToken.None);
await WaitUntilAsync(() => artifactStore.Agenda == "Review API shape");
Assert.Equal("Architecture Sync", noteStore.SavedNote?.Frontmatter.Title);
Assert.Equal(["Ada <ada@example.com>", "Grace"], noteStore.SavedNote?.Frontmatter.Attendees);
@@ -125,6 +143,101 @@ public sealed class RecordingCoordinatorTests
await coordinator.StopAsync(CancellationToken.None);
}
[Fact]
public async Task StartDoesNotWaitForSlowMeetingMetadataButAppliesItLater()
{
var audioSource = new ControlledAudioSource();
var noteStore = new InMemoryMeetingNoteStore("C:\\Vault\\Meetings\\Notes\\metadata-meeting.md");
var artifactStore = new InMemoryMeetingArtifactStore();
var provider = new BlockingMeetingMetadataProvider(new MeetingMetadata(
"Late Calendar Match",
["Ada"],
"Late agenda",
DateTimeOffset.Parse("2026-05-19T12:00:00+02:00")));
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(new EchoStreamingTranscriptionProvider()),
new InMemoryTranscriptStore(),
noteStore,
new CapturingMeetingNoteOpener(),
artifactStore,
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions()),
NullLogger<MeetingRecordingCoordinator>.Instance,
provider);
await coordinator.StartAsync(CancellationToken.None);
Assert.StartsWith("Meeting ", noteStore.SavedNote?.Frontmatter.Title, StringComparison.Ordinal);
provider.Release();
await WaitUntilAsync(() =>
noteStore.SavedNote?.Frontmatter.Title == "Late Calendar Match" &&
artifactStore.Agenda == "Late agenda");
Assert.Equal(["Ada"], noteStore.SavedNote?.Frontmatter.Attendees);
Assert.Equal("Late agenda", artifactStore.Agenda);
Assert.Equal(DateTimeOffset.Parse("2026-05-19T12:00:00+02:00"), artifactStore.ScheduledEnd);
await coordinator.StopAsync(CancellationToken.None);
}
[Fact]
public async Task StartCreatesMeetingNoteWhenMetadataProviderFails()
{
var audioSource = new ControlledAudioSource();
var noteStore = new InMemoryMeetingNoteStore("C:\\Vault\\Meetings\\Notes\\fallback-meeting.md");
var artifactStore = new InMemoryMeetingArtifactStore();
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(new EchoStreamingTranscriptionProvider()),
new InMemoryTranscriptStore("C:\\Vault\\Meetings\\Transcripts\\fallback-transcript.md"),
noteStore,
new CapturingMeetingNoteOpener(),
artifactStore,
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions()),
NullLogger<MeetingRecordingCoordinator>.Instance,
new ThrowingMeetingMetadataProvider());
var status = await coordinator.StartAsync(CancellationToken.None);
Assert.True(status.IsRecording);
Assert.Equal("C:\\Vault\\Meetings\\Notes\\fallback-meeting.md", status.MeetingNotePath);
Assert.NotNull(artifactStore.CreatedArtifacts);
Assert.StartsWith("Meeting ", noteStore.SavedNote?.Frontmatter.Title, StringComparison.Ordinal);
await coordinator.StopAsync(CancellationToken.None);
}
[Fact]
public async Task StartContinuesWhenObsidianOpenFails()
{
var audioSource = new ControlledAudioSource();
var noteStore = new InMemoryMeetingNoteStore("C:\\Vault\\Meetings\\Notes\\open-fails.md");
var artifactStore = new InMemoryMeetingArtifactStore();
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(new EchoStreamingTranscriptionProvider()),
new InMemoryTranscriptStore("C:\\Vault\\Meetings\\Transcripts\\open-fails-transcript.md"),
noteStore,
new ThrowingMeetingNoteOpener(),
artifactStore,
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions()),
NullLogger<MeetingRecordingCoordinator>.Instance);
var status = await coordinator.StartAsync(CancellationToken.None);
Assert.True(status.IsRecording);
Assert.Equal("C:\\Vault\\Meetings\\Notes\\open-fails.md", status.MeetingNotePath);
Assert.NotNull(artifactStore.CreatedArtifacts);
await coordinator.StopAsync(CancellationToken.None);
}
[Fact]
public async Task StopCompletesAudioCaptureAndDrainsFinalTranscriptionWindow()
{
@@ -187,6 +300,33 @@ public sealed class RecordingCoordinatorTests
await coordinator.StopAsync(CancellationToken.None);
}
[Fact]
public async Task StartPassesDictationWordsToSpeechPipeline()
{
var audioSource = new ControlledAudioSource();
var provider = new CapturingPipelineOptionsProvider();
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(provider),
new InMemoryTranscriptStore(),
new InMemoryMeetingNoteStore(),
new CapturingMeetingNoteOpener(),
new InMemoryMeetingArtifactStore(),
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions()),
NullLogger<MeetingRecordingCoordinator>.Instance,
dictationWordStore: new FixedDictationWordStore(["PBI", "Product Backlog Item"]));
await coordinator.StartAsync(CancellationToken.None);
await audioSource.WriteAsync(new AudioChunk([1, 0], 16000, 1), CancellationToken.None);
await provider.OptionsObserved.Task.WaitAsync(TimeSpan.FromSeconds(5));
Assert.Equal(["PBI", "Product Backlog Item"], provider.Options?.DictationWords);
await coordinator.StopAsync(CancellationToken.None);
}
[Fact]
public async Task StopRewritesTranscriptWithFinalDiarizedSegmentsWhenAvailable()
{
@@ -236,6 +376,272 @@ public sealed class RecordingCoordinatorTests
});
}
[Fact]
public async Task StopRelabelsFinishedTranscriptWithLearnedSpeakerIdentity()
{
var audioSource = new CapturedChunkThenCancelAudioSource(new AudioChunk(new byte[] { 1, 0, 2, 0 }, 16000, 1));
var transcriptStore = new InMemoryTranscriptStore();
var noteStore = new InMemoryMeetingNoteStore();
noteStore.UpdateSavedNote(MeetingNoteTemplate.Create(
"Identity Test",
DateTimeOffset.Parse("2026-05-20T12:00:00+02:00"),
attendees: ["Chris"],
transcriptPath: "memory-transcript.md",
assistantContextPath: "memory-context.md",
summaryPath: "memory-summary.md"));
var finalizer = new CapturingTranscriptFinalizer(
[
new TranscriptionSegment(TimeSpan.Zero, TimeSpan.FromSeconds(1), "Guest03", "hello")
]);
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(new FinalSegmentOnAudioCompletionProvider(), finalizer.FinalizeAsync),
transcriptStore,
noteStore,
new CapturingMeetingNoteOpener(),
new InMemoryMeetingArtifactStore(),
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions()),
NullLogger<MeetingRecordingCoordinator>.Instance,
speakerIdentificationService: new FixedSpeakerIdentificationService("Guest03", "Chris"));
await coordinator.StartAsync(CancellationToken.None);
await audioSource.WaitUntilCapturedAsync();
await coordinator.StopAsync(CancellationToken.None);
Assert.Equal("Chris", transcriptStore.ReplacedSegments.Single().Speaker);
}
[Fact]
public async Task StopAddsIdentifiedSpeakerToMeetingAttendees()
{
var audioSource = new CapturedChunkThenCancelAudioSource(new AudioChunk(new byte[] { 1, 0, 2, 0 }, 16000, 1));
var noteStore = new InMemoryMeetingNoteStore();
var finalizer = new CapturingTranscriptFinalizer(
[
new TranscriptionSegment(TimeSpan.Zero, TimeSpan.FromSeconds(1), "Guest03", "hello")
]);
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(new FinalSegmentOnAudioCompletionProvider(), finalizer.FinalizeAsync),
new InMemoryTranscriptStore(),
noteStore,
new CapturingMeetingNoteOpener(),
new InMemoryMeetingArtifactStore(),
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions()),
NullLogger<MeetingRecordingCoordinator>.Instance,
speakerIdentificationService: new FixedSpeakerIdentificationService(
"Guest03",
"Chris",
["Chris", "Christopher"]));
await coordinator.StartAsync(CancellationToken.None);
await audioSource.WaitUntilCapturedAsync();
await coordinator.StopAsync(CancellationToken.None);
var attendees = Assert.IsType<List<string>>(noteStore.SavedNote?.Frontmatter.Attendees);
Assert.Contains("Chris", attendees);
}
[Fact]
public async Task StopDoesNotDuplicateIdentifiedSpeakerWhenAliasIsAlreadyAttendee()
{
var audioSource = new CapturedChunkThenCancelAudioSource(new AudioChunk(new byte[] { 1, 0, 2, 0 }, 16000, 1));
var noteStore = new InMemoryMeetingNoteStore();
var finalizer = new CapturingTranscriptFinalizer(
[
new TranscriptionSegment(TimeSpan.Zero, TimeSpan.FromSeconds(1), "Guest03", "hello")
]);
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(new FinalSegmentOnAudioCompletionProvider(), finalizer.FinalizeAsync),
new InMemoryTranscriptStore(),
noteStore,
new CapturingMeetingNoteOpener(),
new InMemoryMeetingArtifactStore(),
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions()),
NullLogger<MeetingRecordingCoordinator>.Instance,
speakerIdentificationService: new FixedSpeakerIdentificationService(
"Guest03",
"Christopher",
["Christopher", "Chris"]));
await coordinator.StartAsync(CancellationToken.None);
noteStore.UpdateAttendees(["Chris <chris@example.com>"]);
await audioSource.WaitUntilCapturedAsync();
await coordinator.StopAsync(CancellationToken.None);
var attendees = Assert.IsType<List<string>>(noteStore.SavedNote?.Frontmatter.Attendees);
Assert.Equal(["Chris <chris@example.com>"], attendees);
}
[Fact]
public async Task LiveSpeakerIdentityMatchRelabelsCurrentAndFutureTranscriptSegments()
{
var audioSource = new ControlledAudioSource();
var transcriptStore = new InMemoryTranscriptStore();
var speakerIdentification = new BlockingSpeakerIdentificationService("Guest03", "Chris");
var provider = new OrderedChunkProvider();
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(provider),
transcriptStore,
new InMemoryMeetingNoteStore(),
new CapturingMeetingNoteOpener(),
new InMemoryMeetingArtifactStore(),
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions
{
SpeakerIdentification = new SpeakerIdentificationOptions
{
InitialDelay = TimeSpan.Zero,
Interval = TimeSpan.FromMilliseconds(10)
}
}),
NullLogger<MeetingRecordingCoordinator>.Instance,
speakerIdentificationService: speakerIdentification);
await coordinator.StartAsync(CancellationToken.None);
await audioSource.WriteAsync(new AudioChunk(Samples(0, 1, 2, 3, 4, 5, 6, 7), 4, 1), CancellationToken.None);
await WaitUntilAsync(() => transcriptStore.Segments.Any(segment => segment.Text.Contains("first", StringComparison.Ordinal)));
await speakerIdentification.IdentificationObserved.Task.WaitAsync(TimeSpan.FromSeconds(5));
await WaitUntilAsync(() => transcriptStore.ReplacedSegments.Any(segment => segment.Text.Contains("first", StringComparison.Ordinal)));
Assert.Equal("Chris", transcriptStore.ReplacedSegments.Single().Speaker);
await Task.Delay(50);
await audioSource.WriteAsync(new AudioChunk(Samples(8, 9, 10, 11, 12, 13, 14, 15), 4, 1), CancellationToken.None);
await transcriptStore.WaitForTextAsync("second");
await coordinator.StopAsync(CancellationToken.None);
Assert.Collection(
transcriptStore.Segments,
first => Assert.Equal("Guest03", first.Speaker),
second => Assert.Equal("Chris", second.Speaker));
Assert.Contains(
transcriptStore.ReplacedSegments,
segment => segment.Text.Contains("first", StringComparison.Ordinal) && segment.Speaker == "Chris");
}
[Fact]
public async Task LiveSpeakerIdentityMatchAddsIdentifiedSpeakerToMeetingAttendees()
{
var audioSource = new ControlledAudioSource();
var noteStore = new InMemoryMeetingNoteStore();
var speakerIdentification = new BlockingSpeakerIdentificationService("Guest03", "Chris");
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(new OrderedChunkProvider()),
new InMemoryTranscriptStore(),
noteStore,
new CapturingMeetingNoteOpener(),
new InMemoryMeetingArtifactStore(),
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions
{
SpeakerIdentification = new SpeakerIdentificationOptions
{
InitialDelay = TimeSpan.Zero,
Interval = TimeSpan.FromMilliseconds(10)
}
}),
NullLogger<MeetingRecordingCoordinator>.Instance,
speakerIdentificationService: speakerIdentification);
await coordinator.StartAsync(CancellationToken.None);
await audioSource.WriteAsync(new AudioChunk(Samples(0, 1, 2, 3, 4, 5, 6, 7), 4, 1), CancellationToken.None);
await WaitUntilAsync(() => noteStore.SavedNote?.Frontmatter.Attendees.Contains("Chris") == true);
var attendees = Assert.IsType<List<string>>(noteStore.SavedNote?.Frontmatter.Attendees);
Assert.Contains("Chris", attendees);
await coordinator.StopAsync(CancellationToken.None);
}
[Fact]
public async Task LiveSpeakerIdentificationRetriesWhenAttendeesChange()
{
var audioSource = new ControlledAudioSource();
var noteStore = new InMemoryMeetingNoteStore();
var speakerIdentification = new CountingSpeakerIdentificationService();
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(new OrderedChunkProvider()),
new InMemoryTranscriptStore(),
noteStore,
new CapturingMeetingNoteOpener(),
new InMemoryMeetingArtifactStore(),
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions
{
SpeakerIdentification = new SpeakerIdentificationOptions
{
InitialDelay = TimeSpan.Zero,
Interval = TimeSpan.FromMilliseconds(20)
}
}),
NullLogger<MeetingRecordingCoordinator>.Instance,
speakerIdentificationService: speakerIdentification);
await coordinator.StartAsync(CancellationToken.None);
await audioSource.WriteAsync(new AudioChunk(Samples(0, 1, 2, 3, 4, 5, 6, 7), 4, 1), CancellationToken.None);
await WaitUntilAsync(() => speakerIdentification.Requests.Count == 1);
noteStore.UpdateAttendees(["Chris"]);
await WaitUntilAsync(() => speakerIdentification.Requests.Count == 2);
await coordinator.StopAsync(CancellationToken.None);
Assert.Equal(["Chris"], speakerIdentification.Requests.Last().MeetingNote.Frontmatter.Attendees);
}
[Fact]
public async Task LiveSpeakerIdentificationRetriesWhenNewUnmappedSpeakerAppears()
{
var audioSource = new ControlledAudioSource();
var speakerIdentification = new CountingSpeakerIdentificationService();
var coordinator = new MeetingRecordingCoordinator(
audioSource,
new TestSpeechRecognitionPipelineFactory(new ChangingSpeakerOrderedChunkProvider()),
new InMemoryTranscriptStore(),
new InMemoryMeetingNoteStore(),
new CapturingMeetingNoteOpener(),
new InMemoryMeetingArtifactStore(),
new InMemoryRecordedAudioStore(),
new CapturingMeetingSummaryPipeline(),
Options.Create(new MeetingAssistantOptions
{
SpeakerIdentification = new SpeakerIdentificationOptions
{
InitialDelay = TimeSpan.Zero,
Interval = TimeSpan.FromMilliseconds(20)
}
}),
NullLogger<MeetingRecordingCoordinator>.Instance,
speakerIdentificationService: speakerIdentification);
await coordinator.StartAsync(CancellationToken.None);
await audioSource.WriteAsync(new AudioChunk(Samples(0, 1, 2, 3, 4, 5, 6, 7), 4, 1), CancellationToken.None);
await WaitUntilAsync(() => speakerIdentification.Requests.Count == 1);
await audioSource.WriteAsync(new AudioChunk(Samples(8, 9, 10, 11, 12, 13, 14, 15), 4, 1), CancellationToken.None);
await WaitUntilAsync(() => speakerIdentification.Requests.Count == 2);
await coordinator.StopAsync(CancellationToken.None);
var samples = Assert.IsAssignableFrom<IReadOnlyList<SpeakerAudioSample>>(
speakerIdentification.Requests.Last().Samples);
Assert.Contains(samples, sample => sample.Speaker == "Guest04");
}
[Theory]
[InlineData(0, null)]
[InlineData(1, null)]
@@ -538,6 +944,8 @@ public sealed class RecordingCoordinatorTests
public IReadOnlyList<TranscriptionSegment> ReplacedSegments { get; private set; } = [];
public IReadOnlyList<TranscriptionSegment> Segments => segments;
public MeetingNote? MetadataMeetingNote { get; private set; }
public Task ReplaceAsync(
@@ -591,6 +999,11 @@ public sealed class RecordingCoordinatorTests
SavedNote.Frontmatter.Attendees = attendees.ToList();
}
public void UpdateSavedNote(MeetingNote note)
{
SavedNote = note with { Path = notePath };
}
}
private sealed class CapturingMeetingNoteOpener : IMeetingNoteOpener
@@ -634,6 +1047,17 @@ public sealed class RecordingCoordinatorTests
States.Add(state);
return Task.CompletedTask;
}
public Task UpdateAssistantContextMetadataAsync(
MeetingSessionArtifacts artifacts,
string agenda,
DateTimeOffset? scheduledEnd,
CancellationToken cancellationToken)
{
Agenda = agenda;
ScheduledEnd = scheduledEnd;
return Task.CompletedTask;
}
}
private sealed class FixedMeetingMetadataProvider : IMeetingMetadataProvider
@@ -653,6 +1077,48 @@ public sealed class RecordingCoordinatorTests
}
}
private sealed class ThrowingMeetingMetadataProvider : IMeetingMetadataProvider
{
public Task<MeetingMetadata?> GetCurrentMeetingAsync(
DateTimeOffset startedAt,
CancellationToken cancellationToken)
{
throw new InvalidOperationException("metadata unavailable");
}
}
private sealed class BlockingMeetingMetadataProvider : IMeetingMetadataProvider
{
private readonly MeetingMetadata metadata;
private readonly TaskCompletionSource release = new(TaskCreationOptions.RunContinuationsAsynchronously);
public BlockingMeetingMetadataProvider(MeetingMetadata metadata)
{
this.metadata = metadata;
}
public void Release()
{
release.TrySetResult();
}
public async Task<MeetingMetadata?> GetCurrentMeetingAsync(
DateTimeOffset startedAt,
CancellationToken cancellationToken)
{
await release.Task.WaitAsync(cancellationToken);
return metadata;
}
}
private sealed class ThrowingMeetingNoteOpener : IMeetingNoteOpener
{
public Task OpenAsync(string notePath, CancellationToken cancellationToken)
{
throw new InvalidOperationException("obsidian unavailable");
}
}
private sealed class CapturingMeetingSummaryPipeline : IMeetingSummaryPipeline
{
private readonly bool succeeded;
@@ -817,6 +1283,181 @@ public sealed class RecordingCoordinatorTests
}
}
private sealed class FixedSpeakerIdentificationService : ISpeakerIdentificationService
{
private readonly string sourceSpeaker;
private readonly string targetSpeaker;
private readonly IReadOnlyList<string> acceptedNames;
public FixedSpeakerIdentificationService(
string sourceSpeaker,
string targetSpeaker,
IReadOnlyList<string>? acceptedNames = null)
{
this.sourceSpeaker = sourceSpeaker;
this.targetSpeaker = targetSpeaker;
this.acceptedNames = acceptedNames ?? [targetSpeaker];
}
public Task<SpeakerIdentificationResult> ProcessFinishedTranscriptAsync(
SpeakerIdentificationRequest request,
CancellationToken cancellationToken)
{
return Task.FromResult(new SpeakerIdentificationResult(
request.Segments.Select(segment => segment.Speaker == sourceSpeaker
? segment with { Speaker = targetSpeaker }
: segment).ToList(),
new Dictionary<string, string> { [sourceSpeaker] = targetSpeaker },
[new SpeakerIdentityAttendeeMatch(targetSpeaker, acceptedNames)]));
}
public Task<SpeakerIdentificationResult> IdentifyKnownSpeakersAsync(
SpeakerIdentificationRequest request,
CancellationToken cancellationToken)
{
return ProcessFinishedTranscriptAsync(request, cancellationToken);
}
}
private sealed class BlockingSpeakerIdentificationService : ISpeakerIdentificationService
{
private readonly string sourceSpeaker;
private readonly string targetSpeaker;
private readonly IReadOnlyList<string> acceptedNames;
public BlockingSpeakerIdentificationService(
string sourceSpeaker,
string targetSpeaker,
IReadOnlyList<string>? acceptedNames = null)
{
this.sourceSpeaker = sourceSpeaker;
this.targetSpeaker = targetSpeaker;
this.acceptedNames = acceptedNames ?? [targetSpeaker];
}
public TaskCompletionSource IdentificationObserved { get; } =
new(TaskCreationOptions.RunContinuationsAsynchronously);
public Task<SpeakerIdentificationResult> IdentifyKnownSpeakersAsync(
SpeakerIdentificationRequest request,
CancellationToken cancellationToken)
{
IdentificationObserved.TrySetResult();
return Task.FromResult(new SpeakerIdentificationResult(
request.Segments,
new Dictionary<string, string> { [sourceSpeaker] = targetSpeaker },
[new SpeakerIdentityAttendeeMatch(targetSpeaker, acceptedNames)]));
}
public Task<SpeakerIdentificationResult> ProcessFinishedTranscriptAsync(
SpeakerIdentificationRequest request,
CancellationToken cancellationToken)
{
return Task.FromResult(new SpeakerIdentificationResult(request.Segments, new Dictionary<string, string>()));
}
}
private sealed class CountingSpeakerIdentificationService : ISpeakerIdentificationService
{
public List<SpeakerIdentificationRequest> Requests { get; } = [];
public Task<SpeakerIdentificationResult> IdentifyKnownSpeakersAsync(
SpeakerIdentificationRequest request,
CancellationToken cancellationToken)
{
Requests.Add(request);
return Task.FromResult(new SpeakerIdentificationResult(request.Segments, new Dictionary<string, string>()));
}
public Task<SpeakerIdentificationResult> ProcessFinishedTranscriptAsync(
SpeakerIdentificationRequest request,
CancellationToken cancellationToken)
{
return Task.FromResult(new SpeakerIdentificationResult(request.Segments, new Dictionary<string, string>()));
}
}
private sealed class FixedDictationWordStore : IDictationWordStore
{
private readonly IReadOnlyList<string> words;
public FixedDictationWordStore(IReadOnlyList<string> words)
{
this.words = words;
}
public Task<IReadOnlyList<string>> ReadWordsAsync(CancellationToken cancellationToken)
{
return Task.FromResult(words);
}
public Task<IReadOnlyList<string>> AddWordAsync(string word, CancellationToken cancellationToken)
{
return Task.FromResult(words);
}
}
private sealed class CapturingPipelineOptionsProvider : IStreamingTranscriptionProvider
{
public TaskCompletionSource OptionsObserved { get; } =
new(TaskCreationOptions.RunContinuationsAsynchronously);
public SpeechRecognitionPipelineOptions? Options { get; private set; }
public async IAsyncEnumerable<TranscriptionSegment> TranscribeAsync(
IAsyncEnumerable<AudioChunk> audio,
SpeechRecognitionPipelineOptions options,
[System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken)
{
Options = options;
OptionsObserved.TrySetResult();
await foreach (var _ in audio.WithCancellation(cancellationToken))
{
yield return new TranscriptionSegment(TimeSpan.Zero, TimeSpan.Zero, "Unknown", "ok");
}
}
}
private sealed class OrderedChunkProvider : IStreamingTranscriptionProvider
{
public async IAsyncEnumerable<TranscriptionSegment> TranscribeAsync(
IAsyncEnumerable<AudioChunk> audio,
SpeechRecognitionPipelineOptions options,
[System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken)
{
var index = 0;
await foreach (var _ in audio.WithCancellation(cancellationToken))
{
var text = index++ == 0 ? "first" : "second";
yield return new TranscriptionSegment(
TimeSpan.FromSeconds(index - 1),
TimeSpan.FromSeconds(index + 1),
"Guest03",
$"{text} has enough words for identification.");
}
}
}
private sealed class ChangingSpeakerOrderedChunkProvider : IStreamingTranscriptionProvider
{
public async IAsyncEnumerable<TranscriptionSegment> TranscribeAsync(
IAsyncEnumerable<AudioChunk> audio,
SpeechRecognitionPipelineOptions options,
[System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken)
{
var index = 0;
await foreach (var _ in audio.WithCancellation(cancellationToken))
{
var speaker = index++ == 0 ? "Guest03" : "Guest04";
yield return new TranscriptionSegment(
TimeSpan.FromSeconds(index - 1),
TimeSpan.FromSeconds(index + 1),
speaker,
"This segment has enough words for identification.");
}
}
}
private sealed class ControlledAudioSource : IMeetingAudioSource
{
private readonly Channel<AudioChunk> chunks = Channel.CreateUnbounded<AudioChunk>();
@@ -870,6 +1511,7 @@ public sealed class RecordingCoordinatorTests
public async IAsyncEnumerable<TranscriptionSegment> TranscribeAsync(
IAsyncEnumerable<AudioChunk> audio,
SpeechRecognitionPipelineOptions options,
[System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken)
{
await foreach (var chunk in audio.WithCancellation(cancellationToken))
@@ -884,6 +1526,7 @@ public sealed class RecordingCoordinatorTests
{
public async IAsyncEnumerable<TranscriptionSegment> TranscribeAsync(
IAsyncEnumerable<AudioChunk> audio,
SpeechRecognitionPipelineOptions options,
[System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken)
{
var byteCount = 0;
@@ -915,6 +1558,7 @@ public sealed class RecordingCoordinatorTests
public async IAsyncEnumerable<TranscriptionSegment> TranscribeAsync(
IAsyncEnumerable<AudioChunk> audio,
SpeechRecognitionPipelineOptions options,
[System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken)
{
waitingForBackend.TrySetResult();
@@ -925,6 +1569,14 @@ public sealed class RecordingCoordinatorTests
}
}
}
private static byte[] Samples(params short[] samples)
{
var bytes = new byte[samples.Length * sizeof(short)];
Buffer.BlockCopy(samples, 0, bytes, 0, bytes.Length);
return bytes;
}
}