Fix import error for get_result and modify queue to keep jobs queued when no workers available

vidai/backend.py

@@ -29,6 +29,26 @@ worker_sockets = {}  # type: dict
 pending_results = {}  # msg_id -> result message
 
 
+def get_result(msg_id: str) -> dict:
+    """Poll for result from backend via socket."""
+    import uuid
+    from . import comm
+
+    try:
+        # Send get_result request
+        result_msg = Message('get_result', str(uuid.uuid4()), {'request_id': msg_id})
+        comm.send_message(result_msg)
+
+        # Wait for response
+        response = comm.receive_message(timeout=5)
+        if response:
+            return response.data
+        else:
+            return {'error': 'Timeout waiting for result'}
+    except Exception as e:
+        return {'error': str(e)}
+
+
 def handle_web_message(message: Message) -> Message:
     """Handle messages from web interface."""
     if message.msg_type == 'analyze_request':

vidai/queue.py

@@ -88,6 +88,41 @@ class QueueManager:
 
         return True
 
+    def _can_start_job(self, job: Dict[str, Any]) -> bool:
+        """Check if a job can be started (worker available)."""
+        from .cluster_master import cluster_master
+        from .config import get_analysis_backend, get_training_backend
+        from .backend import worker_sockets
+
+        request_type = job['request_type']
+        if request_type == 'analyze':
+            process_type = 'analysis'
+        elif request_type == 'train':
+            process_type = 'training'
+        else:
+            process_type = request_type
+        model_path = job['data'].get('model_path', 'Qwen/Qwen2.5-VL-7B-Instruct')
+
+        # Check for distributed worker
+        worker_key = cluster_master.select_worker_for_job(process_type, model_path, job['data'])
+        if worker_key:
+            return True
+
+        # Check for local worker
+        if request_type == 'analyze':
+            backend = get_analysis_backend()
+            worker_key_local = f'analysis_{backend}'
+        elif request_type == 'train':
+            backend = get_training_backend()
+            worker_key_local = f'training_{backend}'
+        else:
+            return False
+
+        if worker_key_local in worker_sockets:
+            return True
+
+        return False
+
     def _process_queue(self) -> None:
         """Background thread to process queued jobs."""
         while self.running:
@@ -97,6 +132,7 @@ class QueueManager:
                         pending = get_pending_queue_items()
                         if pending:
                             job = pending[0]  # Get highest priority job
+                            if self._can_start_job(job):
                                 self._start_job(job)
 
                 time.sleep(1)  # Check every second