new features

cpv3/modules/transcription/repository.py

@@ -32,6 +32,14 @@ class TranscriptionRepository:
         )
         return result.scalar_one_or_none()
 
+    async def get_by_artifact_id(self, artifact_id: uuid.UUID) -> Transcription | None:
+        result = await self._session.execute(
+            select(Transcription)
+            .where(Transcription.artifact_id == artifact_id)
+            .where(Transcription.is_active.is_(True))
+        )
+        return result.scalar_one_or_none()
+
     async def create(self, data: TranscriptionCreate) -> Transcription:
         transcription = Transcription(
             project_id=data.project_id,

cpv3/modules/transcription/router.py

@@ -67,6 +67,21 @@ async def retrieve_transcription_entry(
     return TranscriptionRead.model_validate(transcription)
 
 
+@router.get("/transcriptions/by-artifact/{artifact_id}/", response_model=TranscriptionRead)
+async def retrieve_transcription_by_artifact(
+    artifact_id: uuid.UUID,
+    current_user: User = Depends(get_current_user),
+    db: AsyncSession = Depends(get_db),
+) -> TranscriptionRead:
+    _ = current_user
+    repo = TranscriptionRepository(db)
+    transcription = await repo.get_by_artifact_id(artifact_id)
+    if transcription is None:
+        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="Not found")
+
+    return TranscriptionRead.model_validate(transcription)
+
+
 @router.patch("/transcriptions/{transcription_id}/", response_model=TranscriptionRead)
 async def patch_transcription_entry(
     transcription_id: uuid.UUID,

cpv3/modules/transcription/service.py

@@ -240,11 +240,15 @@ def _make_document_from_segments(
     return Document(segments=result_segments)
 
 
+ProgressCallback = Callable[[float], None]
+
+
 def _whisper_transcribe_sync(
     *,
     local_file_path: str,
     model_name: str,
     language: str | None,
+    on_progress: ProgressCallback | None = None,
 ) -> Document:
     import whisper  # type: ignore[import-untyped]
 
@@ -267,14 +271,35 @@ def _whisper_transcribe_sync(
         probs = cast(dict[str, float], probs_raw)
         language = max(probs, key=lambda k: probs[k])
 
-    result = whisper.transcribe(
-        audio=whisper.load_audio(local_file_path),
-        model=model,
-        word_timestamps=True,
-        temperature=0.2,
-        language=language,
-        verbose=False,
-    )
+    if on_progress is not None:
+        from unittest.mock import patch
+        from tqdm import tqdm as _orig_tqdm
+
+        class _ProgressTqdm(_orig_tqdm):
+            def update(self, n=1):
+                super().update(n)
+                if self.total:
+                    on_progress(min(self.n / self.total * 100.0, 100.0))
+
+        with patch("whisper.transcribe.tqdm.tqdm", _ProgressTqdm):
+            result = whisper.transcribe(
+                audio=whisper.load_audio(local_file_path),
+                model=model,
+                word_timestamps=True,
+                temperature=0.2,
+                language=language,
+                verbose=False,
+            )
+        on_progress(100.0)
+    else:
+        result = whisper.transcribe(
+            audio=whisper.load_audio(local_file_path),
+            model=model,
+            word_timestamps=True,
+            temperature=0.2,
+            language=language,
+            verbose=None,
+        )
 
     parsed = WhisperResult.model_validate(result)
 
@@ -296,6 +321,7 @@ async def transcribe_with_whisper(
     file_key: str,
     model_name: str = "tiny",
     language: str | None = None,
+    on_progress: ProgressCallback | None = None,
 ) -> Document:
     tmp = await storage.download_to_temp(file_key)
     try:
@@ -304,6 +330,7 @@ async def transcribe_with_whisper(
                 local_file_path=tmp.path,
                 model_name=model_name,
                 language=language,
+                on_progress=on_progress,
             )
         )
     finally: