[Question]: Sydney does not remember previous messages in the chat? #1128
-
Contact DetailsWhat happened?While the normal Bing AI chat is working well with no bugs at all, the jailbroken Sydney does not seem to remember any of the messages in the chat. In fact, it seems that every time you send a message through the chat with Sydney, she acts like it's a brand new chat every single time, I say that because she would introduce herself in every reply and for a while every single reply she made would change the title of the chat, while normally the title of the chat would only be changed by the first messages in a new conversation. As far as I know, the issue doesn't seem to be the cookie string token, I messed with the .env file to no avail. This problem is similar to another problem some other user had, but this issue is only with Sydney, normal Bing chat doesn't have any problems. I didn't find anything else regarding this type of issue, but if something like this was already posted and solved I apologize, I didn't look further enough then. Steps to Reproduce
She'll always answer a variation of "Of course I do! It's <completely random name or 'user'>, you told me before!" Of course, while also reintroducing herself again and again... What browsers are you seeing the problem on?Chrome Relevant log outputNo response ScreenshotsNo response Code of Conduct
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
Hi @Cascaisdude Thanks for your report. I've had to move it to a discussion because there may not be one underlying issue that we can really tackle. Also be sure to update to the latest commits of the project to assure things are working as intended. Any of the installation guides here can walk you through this: https://docs.librechat.ai/install/docker_compose_install.html Technically, every new message using Sydney mode creates a new conversation. It is always using new conversations to string as much of a jailbroken response together as it can. The output of the LLM cannot be guaranteed as results will vary. If something is teetering its "memory," one or all of these conversations are triggering a blocker, whether it's a censor, IP filter, rate limit, etc. As you can see, I cannot reproduce what you reported even now: |
Beta Was this translation helpful? Give feedback.
Hi @Cascaisdude
Thanks for your report. I've had to move it to a discussion because there may not be one underlying issue that we can really tackle.
Also be sure to update to the latest commits of the project to assure things are working as intended. Any of the installation guides here can walk you through this: https://docs.librechat.ai/install/docker_compose_install.html
Technically, every new message using Sydney mode creates a new conversation. It is always using new conversations to string as much of a jailbroken response together as it can. The output of the LLM cannot be guaranteed as results will vary. If something is teetering its "memory," one or all of these conversations are t…