.

vllm_backend.py

@@ -25,8 +25,7 @@ class VLLMChatBackend(ChatBackend):
 
         params = SamplingParams(
             temperature=float(request.get("temperature", 0.7)),
-            max_tokens=int(request.get("max_tokens", 512)),
-            stream=False,  # we want full text only
+            max_tokens=int(request.get("max_tokens", 512))
         )
 
         rid = f"chatcmpl-local-{int(time.time())}"