fix: Once the memory is activated, it cannot be deactivated.

- Avoid misleading front-end users into choosing whether to turn off or keep the memory function enabled after selecting it.

api/core/memory/token_buffer_memory.py

@@ -33,6 +33,9 @@ class TokenBufferMemory:
         :param max_token_limit: max token limit
         :param message_limit: message limit
         """
+        if not message_limit or message_limit <= 0:
+            return []
+
         app_record = self.conversation.app
 
         # fetch limited messages, and return reversed
@@ -52,10 +55,7 @@ class TokenBufferMemory:
             .order_by(Message.created_at.desc())
         )
 
-        if message_limit and message_limit > 0:
+        message_limit = min(message_limit, 500)
-            message_limit = min(message_limit, 500)
-        else:
-            message_limit = 500
 
         messages = query.limit(message_limit).all()