Add speaker-attributed meeting transcription

tests/test_bot.py

@@ -70,13 +70,14 @@ def test_build_retry_notice_mentions_reason_and_path():
     assert "`/retry`" in message
 
 
-def test_meeting_recorder_wraps_wave_sink_with_silence_generator(tmp_path, monkeypatch):
+def test_meeting_recorder_wraps_meeting_sink_with_silence_generator(tmp_path, monkeypatch):
     created = {}
 
-    class FakeWaveSink:
-        def __init__(self, output_path):
+    class FakeMeetingSink:
+        def __init__(self, output_path, tracks_dir):
             self.output_path = output_path
-            created["wave"] = self
+            self.tracks_dir = tracks_dir
+            created["meeting"] = self
 
     class FakeSilenceGeneratorSink:
         def __init__(self, destination):
@@ -94,7 +95,7 @@ def test_meeting_recorder_wraps_wave_sink_with_silence_generator(tmp_path, monke
             self.listened_sink = sink
             created["after"] = after
 
-    monkeypatch.setattr(voice.voice_recv, "WaveSink", FakeWaveSink)
+    monkeypatch.setattr(voice, "MeetingSink", FakeMeetingSink)
     monkeypatch.setattr(voice.voice_recv, "SilenceGeneratorSink", FakeSilenceGeneratorSink)
 
     vc = FakeVoiceClient()
@@ -103,8 +104,9 @@ def test_meeting_recorder_wraps_wave_sink_with_silence_generator(tmp_path, monke
 
     asyncio.run(recorder.start(lambda error: None))
 
-    assert created["wave"].output_path == str(output_path)
-    assert created["silence"].destination is created["wave"]
+    assert created["meeting"].output_path == str(output_path)
+    assert created["meeting"].tracks_dir == str(tmp_path / "tracks")
+    assert created["silence"].destination is created["meeting"]
     assert vc.listened_sink is created["silence"]
     assert recorder.recording is True
 
@@ -129,3 +131,72 @@ def test_wait_for_file_ready_accepts_valid_wav(tmp_path):
     ready = asyncio.run(bot.wait_for_file_ready(str(valid), attempts=3, delay=0.001))
 
     assert ready is True
+
+
+def test_load_speaker_tracks_resolves_manifest_tracks_next_to_recording(tmp_path):
+    meeting = tmp_path / "meeting.wav"
+    meeting.write_bytes(b"RIFFdemo")
+    tracks_dir = tmp_path / "tracks"
+    tracks_dir.mkdir()
+    alice = tracks_dir / "track-111.wav"
+    alice.write_bytes(b"alice")
+    missing = tracks_dir / "track-222.wav"
+    (tracks_dir / "tracks.json").write_text(
+        '{"tracks":[{"display_name":"Alice","path":"missing-prefix/track-111.wav"},'
+        '{"display_name":"Bob","path":"' + str(missing) + '"}]}'
+    )
+
+    tracks = bot._load_speaker_tracks(str(meeting))
+
+    assert tracks == [{"display_name": "Alice", "path": str(alice)}]
+
+
+def test_transcribe_recording_uses_speaker_tracks_when_manifest_exists(tmp_path, monkeypatch):
+    meeting = tmp_path / "meeting.wav"
+    meeting.write_bytes(b"RIFFdemo")
+    tracks_dir = tmp_path / "tracks"
+    tracks_dir.mkdir()
+    alice = tracks_dir / "track-111.wav"
+    alice.write_bytes(b"alice")
+    (tracks_dir / "tracks.json").write_text(
+        '{"tracks":[{"display_name":"Alice","path":"' + str(alice) + '"}]}'
+    )
+    calls = []
+
+    async def fake_transcribe_tracks(tracks):
+        calls.append(("tracks", tracks))
+        return "[Alice]\nhello"
+
+    async def fake_transcribe(path):
+        calls.append(("single", path))
+        return "single transcript"
+
+    monkeypatch.setattr(bot, "transcribe_tracks", fake_transcribe_tracks)
+    monkeypatch.setattr(bot, "transcribe", fake_transcribe)
+
+    result = asyncio.run(bot._transcribe_recording(str(meeting), guild_id=123))
+
+    assert result == "[Alice]\nhello"
+    assert calls == [("tracks", [{"display_name": "Alice", "path": str(alice)}])]
+
+
+def test_transcribe_recording_falls_back_to_single_file_for_old_recordings(tmp_path, monkeypatch):
+    meeting = tmp_path / "meeting.wav"
+    meeting.write_bytes(b"RIFFdemo")
+    calls = []
+
+    async def fake_transcribe_tracks(tracks):
+        calls.append(("tracks", tracks))
+        return "should not be used"
+
+    async def fake_transcribe(path):
+        calls.append(("single", path))
+        return "single transcript"
+
+    monkeypatch.setattr(bot, "transcribe_tracks", fake_transcribe_tracks)
+    monkeypatch.setattr(bot, "transcribe", fake_transcribe)
+
+    result = asyncio.run(bot._transcribe_recording(str(meeting), guild_id=123))
+
+    assert result == "single transcript"
+    assert calls == [("single", str(meeting))]