electron-rare/meet
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Changed diarization (dict) to Transcription object

Browse Source
This commit is contained in:
leo
2026-03-25 17:49:23 +01:00
parent 7dc7da9109
commit e64b2901f9
4 changed files with 44 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/backend/core/api/viewsets.py
+1
View File
@@ -346,6 +346,7 @@ class RoomViewSet(
):
try:
MetadataCollectorService().start(recording)
logger.info("Started MetadataCollectorService")
except MetadataCollectorException:
logger.warning("Failed to start MetadataCollectorService")
src/summary/summary/core/user_assign.py
+7 -6
View File
@@ -181,12 +181,13 @@ def _build_participant_timelines(
def _build_speaker_timelines(
diarization: dict[str, Any],
transcription: Any,
) -> dict[str, list[Interval]]:
"""Build interval timelines from WhisperX diarization segments."""
"""Build interval timelines from WhisperX transcription segments."""
intervals: dict[str, list[Interval]] = {}
for segment in diarization.get("segments", []):
segments = transcription.segments if hasattr(transcription, "segments") else []
for segment in segments:
speaker = segment.get("speaker")
if speaker is None:
continue
@@ -202,7 +203,7 @@ def _build_speaker_timelines(
def assign_speakers(
metadata: dict[str, Any],
diarization: dict[str, Any],
transcription: Any,
recording_start_datetime: datetime,
overlap_threshold: float = DEFAULT_OVERLAP_THRESHOLD,
) -> AssignmentResult:
@@ -210,7 +211,7 @@ def assign_speakers(
Args:
metadata: User metadata with `events` and `participants`.
diarization: WhisperX JSON output containing `segments`.
transcription: WhisperX Transcription object with a `segments` attribute.
recording_start_datetime: UTC datetime for t=0 reference.
overlap_threshold: Minimum overlap/speaker_duration to accept.
@@ -221,7 +222,7 @@ def assign_speakers(
participant_timelines, participant_names = _build_participant_timelines(
metadata, recording_start_datetime
)
speaker_timelines = _build_speaker_timelines(diarization)
speaker_timelines = _build_speaker_timelines(transcription)
result = AssignmentResult()
src/summary/tests/api/test_api_tasks.py
-1
View File
@@ -31,7 +31,6 @@ class TestTasks:
},
)
print(response)
assert response.status_code == 200
assert response.json() == {"id": "task-id-abc", "message": "Task created"}
src/summary/tests/unit/test_user_assign.py
+36 -28
View File
@@ -1,5 +1,6 @@
"""Tests for the speaker-to-user assignment service."""
from dataclasses import dataclass, field
from datetime import datetime
from summary.core.user_assign import (
@@ -12,6 +13,13 @@ from summary.core.user_assign import (
assign_speakers,
)
@dataclass
class FakeTranscription:
"""Mimics the OpenAI Transcription pydantic model for testing."""
segments: list = field(default_factory=list)
RECORDING_START = datetime.fromisoformat("2026-03-17T15:30:33.000001")
METADATA_SINGLE_USER = {
@@ -55,8 +63,8 @@ METADATA_SINGLE_USER = {
],
}
DIARIZATION_SINGLE_SPEAKER = {
"segments": [
DIARIZATION_SINGLE_SPEAKER = FakeTranscription(
segments=[
{
"start": 1.363,
"end": 3.545,
@@ -70,7 +78,7 @@ DIARIZATION_SINGLE_SPEAKER = {
"speaker": "SPEAKER_00",
},
],
}
)
USER_ID = "da8d39ff-3b1c-4e8d-9a70-c630c9871bcb"
@@ -209,13 +217,13 @@ class TestAssignSpeakers:
],
"participants": [{"participantId": "user-a", "name": "Shared Mic"}],
}
diarization = {
"segments": [
transcription = FakeTranscription(
segments=[
{"start": 1.0, "end": 3.0, "speaker": "SPEAKER_00"},
{"start": 5.0, "end": 7.0, "speaker": "SPEAKER_01"},
],
}
result = assign_speakers(metadata, diarization, RECORDING_START)
)
result = assign_speakers(metadata, transcription, RECORDING_START)
assert len(result.assignments) == 2
pids = {a.participant_id for a in result.assignments}
assert pids == {"user-a"}
@@ -250,13 +258,13 @@ class TestAssignSpeakers:
{"participantId": "user-b", "name": "Bob"},
],
}
diarization = {
"segments": [
transcription = FakeTranscription(
segments=[
{"start": 1.5, "end": 3.5, "speaker": "SPEAKER_00"},
{"start": 5.5, "end": 7.5, "speaker": "SPEAKER_01"},
],
}
result = assign_speakers(metadata, diarization, RECORDING_START)
)
result = assign_speakers(metadata, transcription, RECORDING_START)
assert len(result.assignments) == 2
by_speaker = {a.speaker_label: a for a in result.assignments}
assert by_speaker["SPEAKER_00"].participant_name == "Alice"
@@ -296,14 +304,14 @@ class TestAssignSpeakers:
{"participantId": "user-b", "name": "Bob"},
],
}
diarization = {
"segments": [
transcription = FakeTranscription(
segments=[
{"start": 1.5, "end": 5.5, "speaker": "SPEAKER_00"},
{"start": 4.0, "end": 7.5, "speaker": "SPEAKER_01"},
],
}
)
result = assign_speakers(
metadata, diarization, RECORDING_START, overlap_threshold=0.3
metadata, transcription, RECORDING_START, overlap_threshold=0.3
)
assert len(result.assignments) == 2
by_speaker = {a.speaker_label: a for a in result.assignments}
@@ -328,14 +336,14 @@ class TestAssignSpeakers:
],
"participants": [{"participantId": "user-a", "name": "Brief User"}],
}
diarization = {
"segments": [
transcription = FakeTranscription(
segments=[
{"start": 1.0, "end": 10.0, "speaker": "SPEAKER_00"},
],
}
)
result = assign_speakers(
metadata,
diarization,
transcription,
RECORDING_START,
overlap_threshold=0.5,
)
@@ -359,30 +367,30 @@ class TestAssignSpeakers:
],
"participants": [{"participantId": "user-a", "name": "Early User"}],
}
diarization = {
"segments": [
transcription = FakeTranscription(
segments=[
{"start": 0.0, "end": 3.0, "speaker": "SPEAKER_00"},
],
}
result = assign_speakers(metadata, diarization, RECORDING_START)
)
result = assign_speakers(metadata, transcription, RECORDING_START)
assert len(result.assignments) == 1
assert result.assignments[0].participant_name == "Early User"
def test_empty_diarization(self):
"""No segments produces empty result."""
result = assign_speakers(
METADATA_SINGLE_USER, {"segments": []}, RECORDING_START
METADATA_SINGLE_USER, FakeTranscription(segments=[]), RECORDING_START
)
assert result == AssignmentResult()
def test_segment_without_speaker_ignored(self):
"""Segments missing speaker key are skipped."""
diarization = {
"segments": [
transcription = FakeTranscription(
segments=[
{"start": 1.0, "end": 3.0, "text": "no speaker"},
],
}
result = assign_speakers(METADATA_SINGLE_USER, diarization, RECORDING_START)
)
result = assign_speakers(METADATA_SINGLE_USER, transcription, RECORDING_START)
assert result == AssignmentResult()