diff --git a/server/inference.py b/server/inference.py
index 3864d5f..76792fe 100644
--- a/server/inference.py
+++ b/server/inference.py
@@ -29,10 +29,12 @@ async def connect_to_conversation(ws: WebSocket):
     try:
       conversation: llm.AsyncConversation = llm.cli.load_conversation(conversation_id, async_=True)
     except:
-      await ws.send_denial_response(JSONResponse({
-        "error": "unable to load conversation {}".format(conversation_id)
-      }, status_code=404))
-      return
+      # await ws.send_denial_response(JSONResponse({
+      #   "error": "unable to load conversation {}".format(conversation_id)
+      # }, status_code=404))
+      # return
+      model = llm.get_async_model(ws.query_params.get("model"))
+      conversation = llm.AsyncConversation(model, id=conversation_id)
 
   await ws.accept()