логирование ошибок транскрибации для битых файлов

apiApp/config.py

@@ -40,4 +40,7 @@ WEBHOOK_API_KEY = os.getenv("WEBHOOK_API_KEY", "webhook_secret_key")
 # Auto-restore recognition on startup
 ENABLE_AUTO_RESTORE = os.getenv("ENABLE_AUTO_RESTORE", "true").lower() == "true"
 AUTO_RESTORE_LIMIT = int(os.getenv("AUTO_RESTORE_LIMIT", "100"))  # Максимум файлов для восстановления
-AUTO_RESTORE_DELAY = int(os.getenv("AUTO_RESTORE_DELAY", "5"))  # Задержка перед запуском (секунды)
+AUTO_RESTORE_DELAY = int(os.getenv("AUTO_RESTORE_DELAY", "5"))  # Задержка перед запуском (секунды)
+
+# Recognition retry policy (FileAudioAPI side)
+MAX_RECOGNITION_ATTEMPTS = int(os.getenv("MAX_RECOGNITION_ATTEMPTS", "3"))

apiApp/database/Audio.py

@@ -15,6 +15,10 @@ class Audio(Base):
     duration = Column(Float)
     file_size = Column(Integer)
     sourse = Column(Text, default="internal")
+    recognition_status = Column(Text, default="pending", index=True)  # pending, processing, completed, failed
+    recognition_attempts = Column(Integer, default=0)
+    recognition_last_error = Column(Text, nullable=True)
+    recognition_last_attempt_at = Column(DateTime, nullable=True)
 
     ai_conclusion = relationship("AiConclusion", back_populates="audio", cascade="all, delete-orphan")
 
@@ -26,5 +30,9 @@ class Audio(Base):
             "file_path": self.file_path,
             "duration": self.duration,
             "file_size": self.file_size,
-            "sourse": self.sourse
+            "sourse": self.sourse,
+            "recognition_status": self.recognition_status,
+            "recognition_attempts": self.recognition_attempts,
+            "recognition_last_error": self.recognition_last_error,
+            "recognition_last_attempt_at": self.recognition_last_attempt_at.isoformat() if self.recognition_last_attempt_at else None,
         }

apiApp/routers/ai_conclusion_router.py

@@ -62,6 +62,11 @@ class AiConclusionResponse(BaseModel):
     error: Optional[str] = None
 
 
+class RecognitionFailedRequest(BaseModel):
+    filename: str
+    error: str
+
+
 class ConclusionByFilenameResponse(BaseModel):
     """Заключение по имени файла"""
     filename: str
@@ -154,6 +159,11 @@ async def save_ai_conclusion(request: AiConclusionRequest, db: Session = Depends
         db.commit()
         logger.info(f"✅ Заключение сохранено для {request.filename}")
 
+        # Обновляем статус распознавания у Audio
+        audio.recognition_status = "completed"
+        audio.recognition_last_error = None
+        db.commit()
+
         # Для внешних файлов — отправляем результат клиенту из FileAudioAPI
         if (audio.sourse or "").lower() == "external" and request.callback_url:
             _send_callback(request.callback_url, audio, conclusion_data)
@@ -207,3 +217,25 @@ async def save_ai_conclusion(request: AiConclusionRequest, db: Session = Depends
             status_code=500,
             detail=str(e)
         )
+
+
+@ai_conclusion_router.post("/conclusion/failed", response_model=AiConclusionResponse)
+async def mark_recognition_failed(request: RecognitionFailedRequest, db: Session = Depends(get_db)):
+    """
+    Помечает распознавание как failed для файла (чтобы auto-restore не пытался бесконечно).
+    Используется GigaAM_API при невозможности получить результат.
+    """
+    audio = db.query(Audio).filter(Audio.filename == request.filename).first()
+    if not audio:
+        raise HTTPException(status_code=404, detail=f"Файл не найден: {request.filename}")
+
+    audio.recognition_status = "failed"
+    audio.recognition_last_error = request.error
+    db.commit()
+
+    return AiConclusionResponse(
+        success=True,
+        message="Recognition marked as failed",
+        audio_id=str(audio.id),
+        filename=audio.filename
+    )

apiApp/routers/audio_management_router.py

@@ -13,7 +13,7 @@ from datetime import datetime
 from apiApp.database import get_db
 from apiApp.database.Audio import Audio
 from apiApp.database.AiConclusion import AiConclusion
-from apiApp.config import AUDIOFILES_PATH
+from apiApp.config import AUDIOFILES_PATH, MAX_RECOGNITION_ATTEMPTS
 
 logger = logging.getLogger(__name__)
 audio_management_router = APIRouter()
@@ -32,10 +32,14 @@ def query_audio_without_conclusion(db, limit=None):
         AiConclusion.audio_id == Audio.id
     )
 
+    # Берём только те, которые еще можно/нужно распознавать
     query = db.query(Audio).filter(
         ~subquery
     ).filter(
         Audio.sourse == "internal"
+    ).filter(
+        (Audio.recognition_status.in_(["pending", "processing"])) |
+        ((Audio.recognition_status == "failed") & (Audio.recognition_attempts < MAX_RECOGNITION_ATTEMPTS))
     ).order_by(Audio.index_date.asc())
 
     if limit:
@@ -112,6 +116,10 @@ async def register_audio_file(
         audio.filename = filename
         audio.file_size = file_size
         audio.index_date = datetime.utcnow()
+        audio.recognition_status = "pending"
+        audio.recognition_attempts = 0
+        audio.recognition_last_error = None
+        audio.recognition_last_attempt_at = None
 
         db.add(audio)
         db.commit()
@@ -150,6 +158,11 @@ def process_audio_file(audio_id: str, db: Session):
             return
 
         logger.info(f"🎵 Запуск распознавания для {audio.filename}")
+        audio.recognition_status = "processing"
+        audio.recognition_attempts = (audio.recognition_attempts or 0) + 1
+        audio.recognition_last_attempt_at = datetime.utcnow()
+        audio.recognition_last_error = None
+        db.commit()
 
         # Проверяем что файл существует на диске
         from apiApp.config import AUDIOFILES_PATH
@@ -158,7 +171,9 @@ def process_audio_file(audio_id: str, db: Session):
 
         if not os.path.exists(file_path):
             logger.error(f"❌ Файл не найден на диске в FileAudioAPI: {file_path}")
-            # Помечаем audio как проблемный
+            audio.recognition_status = "failed"
+            audio.recognition_last_error = f"File not found on disk: {file_path}"
+            db.commit()
             return
 
         file_size = os.path.getsize(file_path)
@@ -187,11 +202,20 @@ def process_audio_file(audio_id: str, db: Session):
                 error_detail = response.text
                 logger.error(f"❌ Ошибка запуска распознавания для {audio.filename}: {response.status_code}")
                 logger.error(f"   Detail: {error_detail}")
+                audio.recognition_status = "failed"
+                audio.recognition_last_error = f"GigaAM start failed: {response.status_code} {error_detail}"
+                db.commit()
 
         except requests.exceptions.Timeout:
             logger.error(f"❌ Таймаут при отправке задачи для {audio.filename}")
+            audio.recognition_status = "failed"
+            audio.recognition_last_error = "Timeout when starting recognition in GigaAM"
+            db.commit()
         except requests.exceptions.ConnectionError as e:
             logger.error(f"❌ Ошибка подключения к GigaAM API для {audio.filename}: {e}")
+            audio.recognition_status = "failed"
+            audio.recognition_last_error = f"Connection error when starting recognition in GigaAM: {e}"
+            db.commit()
 
     except Exception as e:
         logger.error(f"❌ Ошибка при обработке {audio_id}: {e}")