Update app.py

app.py

@@ -28,7 +28,7 @@ def start_trivia_game():
         model="llama-3.1-70b-versatile",
         messages=conversation_history,
         temperature=1,
-        max_tokens=128000,
+        max_tokens=8000,
         top_p=1,
         stream=True,
         stop=None,
@@ -49,7 +49,7 @@ def continue_trivia_game(user_response):
     conversation_history.append({"role": "user", "content": user_response})
 
     # Token limit management
-    max_tokens = 128000  # Maximum token limit for the LLM (example value)
+    max_tokens = 8000  # Maximum token limit for the LLM (example value)
     current_tokens = count_tokens(conversation_history)
 
     while current_tokens > max_tokens:
@@ -65,7 +65,7 @@ def continue_trivia_game(user_response):
             model="llama-3.1-70b-versatile",
             messages=conversation_history,
             temperature=1,
-            max_tokens=128000,
+            max_tokens=8000,
             top_p=1,
             stream=True,
             stop=None,