Archive summary refinements and profile switching
PR and Push Build/Test / build-and-test (push) Successful in 6m37s

This commit is contained in:
2026-05-27 23:11:21 +02:00
parent c12febe029
commit 7777b349a5
37 changed files with 3190 additions and 121 deletions
@@ -59,6 +59,29 @@ public interface ISpeakerIdentificationService
SpeakerIdentificationRequest request,
CancellationToken cancellationToken);
Task<SpeakerIdentificationResult> IdentifyFinishedSpeakersAsync(
SpeakerIdentificationRequest request,
CancellationToken cancellationToken)
{
return IdentifyKnownSpeakersAsync(request, cancellationToken);
}
Task ApplySpeakerOverrideAsync(
SpeakerIdentificationRequest request,
string sourceSpeaker,
string targetSpeaker,
CancellationToken cancellationToken)
{
return Task.CompletedTask;
}
Task DeleteSpeakerIdentityAsync(
string identity,
CancellationToken cancellationToken)
{
return Task.CompletedTask;
}
Task<SpeakerIdentificationResult> ProcessFinishedTranscriptAsync(
SpeakerIdentificationRequest request,
CancellationToken cancellationToken);
@@ -41,6 +41,94 @@ public sealed class SpeakerIdentityService : ISpeakerIdentificationService
return await ProcessTranscriptAsync(request, mode: SpeakerIdentityProcessingMode.LiveReadOnly, cancellationToken);
}
public async Task<SpeakerIdentificationResult> IdentifyFinishedSpeakersAsync(
SpeakerIdentificationRequest request,
CancellationToken cancellationToken)
{
return await ProcessTranscriptAsync(request, mode: SpeakerIdentityProcessingMode.LiveReadOnly, cancellationToken);
}
public async Task ApplySpeakerOverrideAsync(
SpeakerIdentificationRequest request,
string sourceSpeaker,
string targetSpeaker,
CancellationToken cancellationToken)
{
if (!options.Enabled ||
string.IsNullOrWhiteSpace(sourceSpeaker) ||
string.IsNullOrWhiteSpace(targetSpeaker) ||
string.Equals(sourceSpeaker, targetSpeaker, StringComparison.OrdinalIgnoreCase))
{
return;
}
var sourceLabel = sourceSpeaker.Trim();
var targetName = targetSpeaker.Trim();
await using var context = await dbContextFactory.CreateDbContextAsync(cancellationToken);
await SpeakerIdentitySchema.EnsureCreatedOrUpdatedAsync(context, cancellationToken);
var now = DateTimeOffset.UtcNow;
var meetingReference = CreateReference(request.MeetingNote, now);
var snippet = await ResolveOverrideSnippetAsync(request, sourceLabel, cancellationToken);
var target = await FindIdentityByAcceptedNameAsync(context, targetName, cancellationToken);
var sourceCandidate = await FindCurrentRunCandidateAsync(
context,
meetingReference,
targetName,
cancellationToken);
if (target is null)
{
target = sourceCandidate ?? new SpeakerIdentity
{
CreatedAt = now,
CandidateNames = []
};
if (target.Id == 0)
{
context.SpeakerIdentities.Add(target);
}
}
else if (sourceCandidate is not null && sourceCandidate.Id != target.Id)
{
MergeOverrideCandidate(target, sourceCandidate);
context.SpeakerIdentities.Remove(sourceCandidate);
}
target.CanonicalName = targetName;
target.UpdatedAt = now;
ResetCandidates(target, [targetName]);
AddMeetingReference(target, meetingReference);
AddSnippetIfNeeded(target, snippet);
await context.SaveChangesAsync(cancellationToken);
await SpeakerIdentityTranscriptAudit.AppendIdentifiedAsync(
target.References,
sourceLabel,
targetName,
cancellationToken);
}
public async Task DeleteSpeakerIdentityAsync(
string identity,
CancellationToken cancellationToken)
{
if (!options.Enabled || string.IsNullOrWhiteSpace(identity))
{
return;
}
await using var context = await dbContextFactory.CreateDbContextAsync(cancellationToken);
await SpeakerIdentitySchema.EnsureCreatedOrUpdatedAsync(context, cancellationToken);
var target = await FindIdentityByAcceptedNameAsync(context, identity.Trim(), cancellationToken);
if (target is null)
{
return;
}
context.SpeakerIdentities.Remove(target);
await context.SaveChangesAsync(cancellationToken);
}
private async Task<SpeakerIdentificationResult> ProcessTranscriptAsync(
SpeakerIdentificationRequest request,
SpeakerIdentityProcessingMode mode,
@@ -250,6 +338,115 @@ public sealed class SpeakerIdentityService : ISpeakerIdentificationService
return null;
}
private async Task<byte[]> ResolveOverrideSnippetAsync(
SpeakerIdentificationRequest request,
string sourceSpeaker,
CancellationToken cancellationToken)
{
var sample = request.Samples?
.Where(sample => string.Equals(sample.Speaker, sourceSpeaker, StringComparison.OrdinalIgnoreCase))
.OrderByDescending(sample => sample.Score)
.FirstOrDefault();
if (sample is not null)
{
return sample.WavBytes;
}
var segments = request.Segments
.Where(segment => string.Equals(segment.Speaker, sourceSpeaker, StringComparison.OrdinalIgnoreCase))
.ToList();
return segments.Count == 0
? []
: await snippetExtractor.ExtractSnippetAsync(request.AudioPath, segments, cancellationToken);
}
private static async Task<SpeakerIdentity?> FindIdentityByAcceptedNameAsync(
SpeakerIdentityDbContext context,
string name,
CancellationToken cancellationToken)
{
var identities = await context.SpeakerIdentities
.Include(identity => identity.CandidateNames)
.Include(identity => identity.Aliases)
.Include(identity => identity.Snippets)
.Include(identity => identity.References)
.ToListAsync(cancellationToken);
return identities
.Where(identity => GetAcceptedNames(identity).Contains(name))
.OrderBy(identity => string.Equals(identity.CanonicalName, name, StringComparison.OrdinalIgnoreCase) ? 0 : 1)
.ThenBy(identity => identity.Id)
.FirstOrDefault();
}
private static async Task<SpeakerIdentity?> FindCurrentRunCandidateAsync(
SpeakerIdentityDbContext context,
SpeakerIdentityReference reference,
string targetName,
CancellationToken cancellationToken)
{
var candidates = await context.SpeakerIdentities
.Include(identity => identity.CandidateNames)
.Include(identity => identity.Aliases)
.Include(identity => identity.Snippets)
.Include(identity => identity.References)
.Where(identity => string.IsNullOrWhiteSpace(identity.CanonicalName))
.ToListAsync(cancellationToken);
return candidates
.Where(identity => identity.References.Any(existing => IsSameReference(existing, reference)))
.Where(identity => identity.CandidateNames.Any(candidate =>
string.Equals(candidate.Name, targetName, StringComparison.OrdinalIgnoreCase)) ||
identity.Aliases.Any(alias =>
string.Equals(alias.Name, targetName, StringComparison.OrdinalIgnoreCase)))
.OrderBy(identity => identity.Id)
.FirstOrDefault();
}
private void MergeOverrideCandidate(
SpeakerIdentity target,
SpeakerIdentity source)
{
AddAlias(target, source.CanonicalName);
foreach (var alias in source.Aliases)
{
AddAlias(target, alias.Name);
}
foreach (var candidate in source.CandidateNames)
{
AddAlias(target, candidate.Name);
}
foreach (var reference in source.References)
{
SpeakerIdentityReferences.AddIfMissing(target, reference);
}
foreach (var snippet in source.Snippets)
{
AddSnippetIfNeeded(target, snippet.WavBytes);
}
}
private static void AddAlias(SpeakerIdentity identity, string? alias)
{
if (string.IsNullOrWhiteSpace(alias) ||
string.Equals(identity.CanonicalName, alias, StringComparison.OrdinalIgnoreCase) ||
identity.Aliases.Any(existing => string.Equals(existing.Name, alias, StringComparison.OrdinalIgnoreCase)))
{
return;
}
identity.Aliases.Add(new SpeakerAlias { Name = alias.Trim() });
}
private static bool IsSameReference(
SpeakerIdentityReference first,
SpeakerIdentityReference second)
{
return string.Equals(first.MeetingNotePath, second.MeetingNotePath, StringComparison.OrdinalIgnoreCase) &&
string.Equals(first.TranscriptPath, second.TranscriptPath, StringComparison.OrdinalIgnoreCase);
}
private static bool MatchesAcceptedNames(
SpeakerIdentity identity,
IReadOnlySet<string> names)
@@ -0,0 +1,265 @@
using MeetingAssistant.MeetingNotes;
using System.Text.Json;
using System.Text.RegularExpressions;
namespace MeetingAssistant.Speakers;
public sealed record SpeakerOverride(string From, string To, bool Merge = false);
public sealed record SpeakerIdentityDeletion(string Identity, string Replacement);
public static partial class SpeakerOverrideArtifacts
{
public static async Task ApplyToTranscriptAsync(
string transcriptPath,
SpeakerOverride speakerOverride,
CancellationToken cancellationToken = default)
{
if (!File.Exists(transcriptPath))
{
return;
}
await RewriteTranscriptSpeakerAsync(
transcriptPath,
speakerOverride.From,
speakerOverride.To,
cancellationToken);
}
public static async Task<bool> TranscriptContainsSpeakerAsync(
string transcriptPath,
string speaker,
CancellationToken cancellationToken = default)
{
if (!File.Exists(transcriptPath))
{
return false;
}
var content = await File.ReadAllTextAsync(transcriptPath, cancellationToken);
return TranscriptContainsSpeaker(content, speaker);
}
public static async Task AppendToAssistantContextAsync(
string assistantContextPath,
SpeakerOverride speakerOverride,
CancellationToken cancellationToken = default)
{
await AppendMarkerAsync(assistantContextPath, FormatMarker(speakerOverride), cancellationToken);
}
public static async Task<SpeakerIdentityDeletion> ApplyDeletionToTranscriptAsync(
string transcriptPath,
string identity,
CancellationToken cancellationToken = default)
{
var trimmedIdentity = identity.Trim();
var replacement = "Removed-1";
if (!File.Exists(transcriptPath))
{
return new SpeakerIdentityDeletion(trimmedIdentity, replacement);
}
var content = await File.ReadAllTextAsync(transcriptPath, cancellationToken);
replacement = NextRemovedSpeakerLabel(content);
await RewriteTranscriptSpeakerAsync(
transcriptPath,
trimmedIdentity,
replacement,
content,
cancellationToken);
return new SpeakerIdentityDeletion(trimmedIdentity, replacement);
}
public static async Task AppendIdentityDeletionToAssistantContextAsync(
string assistantContextPath,
SpeakerIdentityDeletion deletion,
CancellationToken cancellationToken = default)
{
await AppendMarkerAsync(assistantContextPath, FormatMarker(deletion), cancellationToken);
}
public static async Task<IReadOnlyList<SpeakerOverride>> ReadFromAssistantContextAsync(
string assistantContextPath,
CancellationToken cancellationToken = default)
{
if (!File.Exists(assistantContextPath))
{
return [];
}
var content = await File.ReadAllTextAsync(assistantContextPath, cancellationToken);
var overrides = new List<SpeakerOverride>();
foreach (Match match in SpeakerOverrideMarkerRegex().Matches(content))
{
SpeakerOverrideJson? parsed;
try
{
parsed = JsonSerializer.Deserialize<SpeakerOverrideJson>(match.Groups["json"].Value);
}
catch (JsonException)
{
continue;
}
if (!string.IsNullOrWhiteSpace(parsed?.From) && !string.IsNullOrWhiteSpace(parsed.To))
{
overrides.Add(new SpeakerOverride(parsed.From.Trim(), parsed.To.Trim(), parsed.Merge));
}
}
return overrides
.DistinctBy(speakerOverride => $"{speakerOverride.From}\n{speakerOverride.To}\n{speakerOverride.Merge}", StringComparer.OrdinalIgnoreCase)
.ToList();
}
public static async Task<IReadOnlyList<SpeakerIdentityDeletion>> ReadIdentityDeletionsFromAssistantContextAsync(
string assistantContextPath,
CancellationToken cancellationToken = default)
{
if (!File.Exists(assistantContextPath))
{
return [];
}
var content = await File.ReadAllTextAsync(assistantContextPath, cancellationToken);
var deletions = new List<SpeakerIdentityDeletion>();
foreach (Match match in SpeakerIdentityDeletionMarkerRegex().Matches(content))
{
SpeakerIdentityDeletionJson? parsed;
try
{
parsed = JsonSerializer.Deserialize<SpeakerIdentityDeletionJson>(match.Groups["json"].Value);
}
catch (JsonException)
{
continue;
}
if (!string.IsNullOrWhiteSpace(parsed?.Identity) && !string.IsNullOrWhiteSpace(parsed.Replacement))
{
deletions.Add(new SpeakerIdentityDeletion(parsed.Identity.Trim(), parsed.Replacement.Trim()));
}
}
return deletions
.DistinctBy(deletion => $"{deletion.Identity}\n{deletion.Replacement}", StringComparer.OrdinalIgnoreCase)
.ToList();
}
private static string FormatMarker(SpeakerOverride speakerOverride)
{
var json = JsonSerializer.Serialize(new SpeakerOverrideJson(
speakerOverride.From,
speakerOverride.To,
speakerOverride.Merge));
return $"<!-- meeting-assistant:speaker-override {json} -->";
}
private static async Task<string> ReadExistingAssistantContextAsync(
string assistantContextPath,
CancellationToken cancellationToken)
{
if (!File.Exists(assistantContextPath))
{
throw new FileNotFoundException(
"Speaker identity markers can only be written to an existing assistant context note.",
assistantContextPath);
}
return await File.ReadAllTextAsync(assistantContextPath, cancellationToken);
}
private static async Task RewriteTranscriptSpeakerAsync(
string transcriptPath,
string from,
string to,
CancellationToken cancellationToken)
{
var content = await File.ReadAllTextAsync(transcriptPath, cancellationToken);
await RewriteTranscriptSpeakerAsync(transcriptPath, from, to, content, cancellationToken);
}
private static async Task RewriteTranscriptSpeakerAsync(
string transcriptPath,
string from,
string to,
string content,
CancellationToken cancellationToken)
{
var pattern = $@"^(\[[^\]]+\]\s*){Regex.Escape(from)}:";
var updated = Regex.Replace(
content,
pattern,
match => $"{match.Groups[1].Value}{to}:",
RegexOptions.Multiline);
if (!string.Equals(content, updated, StringComparison.Ordinal))
{
await File.WriteAllTextAsync(transcriptPath, updated, cancellationToken);
}
}
private static async Task AppendMarkerAsync(
string assistantContextPath,
string marker,
CancellationToken cancellationToken)
{
var content = await ReadExistingAssistantContextAsync(assistantContextPath, cancellationToken);
if (content.Contains(marker, StringComparison.Ordinal))
{
return;
}
var (frontmatter, body) = MeetingArtifactFrontmatterRenderer.Split(content);
var updatedBody = body.TrimEnd('\r', '\n') +
(string.IsNullOrWhiteSpace(body) ? "" : Environment.NewLine + Environment.NewLine) +
marker +
Environment.NewLine;
var updated = string.IsNullOrWhiteSpace(frontmatter)
? updatedBody
: "---" + Environment.NewLine + frontmatter + Environment.NewLine + "---" + Environment.NewLine + Environment.NewLine + updatedBody;
await File.WriteAllTextAsync(assistantContextPath, updated, cancellationToken);
}
private static bool TranscriptContainsSpeaker(string content, string speaker)
{
var pattern = $@"^\[[^\]]+\]\s*{Regex.Escape(speaker.Trim())}:";
return Regex.IsMatch(content, pattern, RegexOptions.IgnoreCase | RegexOptions.Multiline);
}
private static string FormatMarker(SpeakerIdentityDeletion deletion)
{
var json = JsonSerializer.Serialize(new SpeakerIdentityDeletionJson(deletion.Identity, deletion.Replacement));
return $"<!-- meeting-assistant:speaker-identity-deletion {json} -->";
}
private static string NextRemovedSpeakerLabel(string content)
{
var next = RemovedSpeakerLabelRegex()
.Matches(content)
.Select(match => int.TryParse(match.Groups["number"].Value, out var number) ? number : 0)
.DefaultIfEmpty(0)
.Max() + 1;
return $"Removed-{next}";
}
[GeneratedRegex(@"<!--\s*meeting-assistant:speaker-override\s+(?<json>\{.*?\})\s*-->", RegexOptions.IgnoreCase | RegexOptions.Singleline)]
private static partial Regex SpeakerOverrideMarkerRegex();
[GeneratedRegex(@"<!--\s*meeting-assistant:speaker-identity-deletion\s+(?<json>\{.*?\})\s*-->", RegexOptions.IgnoreCase | RegexOptions.Singleline)]
private static partial Regex SpeakerIdentityDeletionMarkerRegex();
[GeneratedRegex(@"^\[[^\]]+\]\s*Removed-(?<number>\d+):", RegexOptions.IgnoreCase | RegexOptions.Multiline)]
private static partial Regex RemovedSpeakerLabelRegex();
private sealed record SpeakerOverrideJson(
[property: System.Text.Json.Serialization.JsonPropertyName("from")] string From,
[property: System.Text.Json.Serialization.JsonPropertyName("to")] string To,
[property: System.Text.Json.Serialization.JsonPropertyName("merge")] bool Merge = false);
private sealed record SpeakerIdentityDeletionJson(
[property: System.Text.Json.Serialization.JsonPropertyName("identity")] string Identity,
[property: System.Text.Json.Serialization.JsonPropertyName("replacement")] string Replacement);
}