Update app.py

app.py

@@ -18,14 +18,17 @@ def respond(message, history: list[tuple[str, str]]):
     messages.append({"role": "user", "content": message})
 
     response = ""
-    for message in client.chat_completion(
+    for token_message in client.chat_completion(
         messages,
         max_tokens=max_tokens,
         stream=True,
         temperature=temperature,
         top_p=top_p,
     ):
-        token = message.choices[0].delta.content
+        # Retrieve token content; if empty, break the loop
+        token = token_message.choices[0].delta.get("content", "")
+        if not token:
+            break
         response += token
         yield response