Update app.py

app.py

@@ -1060,6 +1060,7 @@ def chat_response_stream_multiturn(
     global llm, RES_PRINTED
     assert llm is not None
     assert system_prompt.strip() != '', f'system prompt is empty'
+    tokenizer = llm.get_tokenizer()
     # force removing all 
     vllm_abort(llm)
 
@@ -1068,6 +1069,8 @@ def chat_response_stream_multiturn(
     max_tokens = int(max_tokens)
 
     message = message.strip()
+    if len(message) == 0:
+        raise gr.Error("The message cannot be empty!")
 
     message_safety = safety_check(message, history=history)
     if message_safety is not None:
@@ -1079,6 +1082,9 @@ def chat_response_stream_multiturn(
         message, history, sys_prompt=system_prompt
     )
 
+    if len(tokenizer.encode(full_prompt, add_special_tokens=False)) >= 1000:
+        raise gr.Error(f"Conversation or prompt is too long, please clear the chatbox or try shorter input.")
+
     sampling_params = SamplingParams(
         temperature=temperature, 
         max_tokens=max_tokens,
@@ -1103,7 +1109,7 @@ def chat_response_stream_multiturn(
         item = next(iter(gen.values()))
         cur_out = item.outputs[0].text
         
-    print(f'@@@@@@@@@@\n{full_prompt}<<<{cur_out}>>>\n@@@@@@@@@@\n')
+    print(f'@@@@@@@@@@\n{full_prompt}<<<{cur_out}>>>\n##########\n')
     
     if cur_out is not None and "\\n" in cur_out:
         print(f'double slash-n in cur_out:\n{cur_out}')