Update scripts/chat_web.py

scripts/chat_web.py

@@ -51,10 +51,10 @@ from nanochat.checkpoint_manager import load_model
 from nanochat.engine import Engine
 
 # Victorian system prompt — prepended to first user turn during inference
-SYSTEM_PREFIX = (
-    "[You are a learned Victorian gentleman in conversation. "
-    "Address the question or remark put to you directly.]\n\n"
-)
+# SYSTEM_PREFIX = (
+#    "[You are a learned Victorian gentleman in conversation. "
+#    "Address the question or remark put to you directly.]\n\n"
+# )
 
 # Abuse prevention limits
 MAX_MESSAGES_PER_REQUEST = 500
@@ -399,7 +399,7 @@ async def chat_completions(request: ChatRequest):
                     content = content[0].upper() + content[1:]
                 # Prepend system prompt to the first user turn
                 if turn_count == 0:
-                    content = SYSTEM_PREFIX + content
+                    content = content
                 conversation_tokens.append(user_start)
                 conversation_tokens.extend(worker.tokenizer.encode(content))
                 conversation_tokens.append(user_end)