Refactor inference engine and task management: Remove obsolete inference engine and ONNX engine files, update inference processing to utilize batch handling, and enhance task management structure in documentation. Adjust paths for task specifications to align with new directory organization.

main.py

@@ -22,7 +22,7 @@ ANNOTATIONS_URL = os.environ.get("ANNOTATIONS_URL", "http://annotations:8080")
 loader_client = LoaderHttpClient(LOADER_URL)
 inference = None
 _event_queues: list[asyncio.Queue] = []
-_active_detections: dict[str, bool] = {}
+_active_detections: dict[str, asyncio.Task] = {}
 
 
 class TokenManager:
@@ -109,9 +109,7 @@ class AIConfigDto(BaseModel):
 
 def detection_to_dto(det) -> DetectionDto:
     import constants_inf
-    label = ""
-    if det.cls in constants_inf.annotations_dict:
-        label = constants_inf.annotations_dict[det.cls].name
+    label = constants_inf.get_annotation_name(det.cls)
     return DetectionDto(
         centerX=det.x,
         centerY=det.y,
@@ -197,7 +195,8 @@ def _post_annotation_to_service(token_mgr: TokenManager, media_id: str,
 
 @app.post("/detect/{media_id}")
 async def detect_media(media_id: str, request: Request, config: Optional[AIConfigDto] = None):
-    if media_id in _active_detections:
+    existing = _active_detections.get(media_id)
+    if existing is not None and not existing.done():
         raise HTTPException(status_code=409, detail="Detection already in progress for this media")
 
     auth_header = request.headers.get("authorization", "")
@@ -208,13 +207,19 @@ async def detect_media(media_id: str, request: Request, config: Optional[AIConfi
     cfg = config or AIConfigDto()
     config_dict = cfg.model_dump()
 
-    _active_detections[media_id] = True
-
     async def run_detection():
         loop = asyncio.get_event_loop()
+
+        def _enqueue(event):
+            for q in _event_queues:
+                try:
+                    q.put_nowait(event)
+                except asyncio.QueueFull:
+                    pass
+
         try:
             inf = get_inference()
-            if inf.engine is None:
+            if not inf.is_engine_ready:
                 raise RuntimeError("Detection service unavailable")
 
             def on_annotation(annotation, percent):
@@ -225,12 +230,7 @@ async def detect_media(media_id: str, request: Request, config: Optional[AIConfi
                     mediaStatus="AIProcessing",
                     mediaPercent=percent,
                 )
-                for q in _event_queues:
-                    try:
-                        q.put_nowait(event)
-                    except asyncio.QueueFull:
-                        pass
-
+                loop.call_soon_threadsafe(_enqueue, event)
                 if token_mgr and dtos:
                     _post_annotation_to_service(token_mgr, media_id, annotation, dtos)
 
@@ -241,11 +241,7 @@ async def detect_media(media_id: str, request: Request, config: Optional[AIConfi
                     mediaStatus="AIProcessed",
                     mediaPercent=100,
                 )
-                for q in _event_queues:
-                    try:
-                        q.put_nowait(event)
-                    except asyncio.QueueFull:
-                        pass
+                loop.call_soon_threadsafe(_enqueue, event)
 
             await loop.run_in_executor(
                 executor, inf.run_detect, config_dict, on_annotation, on_status
@@ -257,15 +253,11 @@ async def detect_media(media_id: str, request: Request, config: Optional[AIConfi
                 mediaStatus="Error",
                 mediaPercent=0,
             )
-            for q in _event_queues:
-                try:
-                    q.put_nowait(error_event)
-                except asyncio.QueueFull:
-                    pass
+            _enqueue(error_event)
         finally:
             _active_detections.pop(media_id, None)
 
-    asyncio.create_task(run_detection())
+    _active_detections[media_id] = asyncio.create_task(run_detection())
     return {"status": "started", "mediaId": media_id}