Fix duplicate backend startup in cluster client

- Removed backend process startup from cluster_client.py since vidai.py already starts it for client mode
- This prevents 'Address already in use' error when running as cluster client
- Cluster client now only manages worker processes, not the backend

vidai/cluster_client.py

@@ -129,7 +129,7 @@ class ClusterClient:
                 self.connected = False
 
     async def start_local_processes(self) -> None:
-        """Start local backend and worker processes based on available backends (GPU and CPU)."""
+        """Start local worker processes based on available backends (GPU and CPU)."""
         from .compat import detect_gpu_backends, get_available_backends
 
         gpu_info = detect_gpu_backends()
@@ -140,14 +140,6 @@ class ClusterClient:
 
         # Only start processes if not already started
         if not self.local_processes:
-            # Start local backend process first
-            backend_cmd = [sys.executable, '-m', 'vidai.backend']
-            self.local_processes['backend'] = subprocess.Popen(backend_cmd)
-            print("Started local backend process")
-
-            # Give backend time to start up
-            await asyncio.sleep(2)
-
             # Start analysis workers for available backends (including CPU)
             for backend in available_backends:
                 proc_name = f'analysis_{backend}'
@@ -489,13 +481,10 @@ class ClusterClient:
             except:
                 pass
 
-        # Clear the processes dict (but keep backend)
-        backend_proc = self.local_processes.get('backend')
+        # Clear the processes dict
         self.local_processes.clear()
         self.process_weights.clear()
         self.process_models.clear()
-        if backend_proc:
-            self.local_processes['backend'] = backend_proc
 
         # Check if backend is available
         from .compat import get_available_backends