feat: Celery setup and transcription task — Whisper + FFmpeg pipeline

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/app/tasks/celery_app.py

@@ -0,0 +1,27 @@
+from celery import Celery
+from celery.schedules import crontab
+
+from app.config import settings
+
+celery_app = Celery(
+    "meeting_protocol",
+    broker=settings.redis_url,
+    backend=settings.redis_url,
+)
+
+celery_app.conf.update(
+    task_serializer="json",
+    accept_content=["json"],
+    result_serializer="json",
+    timezone="UTC",
+    enable_utc=True,
+    task_track_started=True,
+    beat_schedule={
+        "cleanup-old-data": {
+            "task": "app.tasks.cleanup.cleanup_old_data",
+            "schedule": crontab(hour=3, minute=0),
+        },
+    },
+)
+
+celery_app.autodiscover_tasks(["app.tasks"])

backend/app/tasks/transcription.py

@@ -0,0 +1,69 @@
+import os
+import uuid
+
+from sqlalchemy import select
+from sqlalchemy.orm import Session
+
+from app.tasks.celery_app import celery_app
+from app.config import settings
+from app.models.recording import Recording, RecordingStatus
+from app.models.transcription import Transcription
+from app.services.transcription import extract_audio, transcribe_audio
+
+from sqlalchemy import create_engine
+from sqlalchemy.orm import sessionmaker
+
+sync_db_url = settings.database_url.replace("+asyncpg", "+psycopg2").replace(
+    "postgresql+psycopg2", "postgresql"
+)
+sync_engine = create_engine(sync_db_url)
+SyncSession = sessionmaker(sync_engine)
+
+
+@celery_app.task(bind=True, name="app.tasks.transcription.transcribe_recording")
+def transcribe_recording(self, recording_id: str):
+    rec_uuid = uuid.UUID(recording_id)
+
+    with SyncSession() as db:
+        recording = db.execute(
+            select(Recording).where(Recording.id == rec_uuid)
+        ).scalar_one_or_none()
+
+        if not recording:
+            return {"error": "Recording not found"}
+
+        recording.status = RecordingStatus.processing
+        db.commit()
+
+        try:
+            file_path = recording.file_path
+            audio_path = file_path
+
+            video_formats = {"mp4", "webm", "mkv", "avi", "mov"}
+            if recording.format in video_formats:
+                audio_path = extract_audio(file_path)
+
+            result = transcribe_audio(audio_path)
+
+            if audio_path != file_path and os.path.exists(audio_path):
+                os.remove(audio_path)
+
+            transcription = Transcription(
+                recording_id=rec_uuid,
+                text=result["text"],
+                segments=result["segments"],
+                language=result["language"],
+                whisper_model=settings.whisper_model,
+                processing_time=result["processing_time"],
+            )
+            db.add(transcription)
+            recording.status = RecordingStatus.done
+            db.commit()
+            db.refresh(transcription)
+
+            return {"transcription_id": str(transcription.id), "status": "done"}
+
+        except Exception as e:
+            recording.status = RecordingStatus.error
+            db.commit()
+            raise self.retry(exc=e, max_retries=1, countdown=30)