limit query length to 500

app.py

@@ -20,17 +20,17 @@ def ask(question, history, behavior):
             {"role":"user" if i%2==0 else "assistant", "content":content}
             for i,content in enumerate(history + [question])
         ]
-        length_messages = num_tokens_from_messages(messages)
-        time_penalty = (length_messages-1000)//10
-        if time_penalty>0:
-            print(f"sleep for {time_penalty:.2f}s for too long a quest: {length_messages}")
-            time.sleep(time_penalty)
-        response = openai.ChatCompletion.create(
-            model="gpt-3.5-turbo",
-            messages=forget_long_term(messages)
-        )["choices"][0]["message"]["content"]
-        while response.startswith("\n"):
-            response = response[1:]
+        raw_length = num_tokens_from_messages(messages)
+        messages=forget_long_term(messages)
+        if len(messages)==0:
+            response = 'Your query is too long and expensive: {raw_length}>500 tokens'
+        else:
+            response = openai.ChatCompletion.create(
+                model="gpt-3.5-turbo",
+                messages=messages
+            )["choices"][0]["message"]["content"]
+            while response.startswith("\n"):
+                response = response[1:]
     except Exception as e:
         print(e)
         response = 'Timeout! Please wait a few minutes and retry'
@@ -57,7 +57,7 @@ def num_tokens_from_messages(messages, model="gpt-3.5-turbo"):
         raise NotImplementedError(f"""num_tokens_from_messages() is not presently implemented for model {model}.
 See https://github.com/openai/openai-python/blob/main/chatml.md for information on how messages are converted to tokens.""")
 
-def forget_long_term(messages, max_num_tokens=4000):
+def forget_long_term(messages, max_num_tokens=500):
     while num_tokens_from_messages(messages)>max_num_tokens:
         if messages[0]["role"]=="system" and not len(messages[0]["content"]>=max_num_tokens):
             messages = messages[:1] + messages[2:]