Here is my user case, I’m building an where users can fetch the accounts associated with their workspace, in order to fetch action needs workspaceid as param, i dont want the user to enter it everytime they have a new chat.
I thought of adding custom instruction to hold the id in memory , but the values die after the chat session is close.
Thing here is if i prompt to save in memory in regular chatgpt , the memory is updated, but not via custom gpt, it tells so but the user memory is not updated.
I recommend implementing a few actions and functions to retrieve and store those variables in your own database, which should be accessible via the API.
Hi,
I’m currently using a Custom GPT to help me learn and improve my English.
One of the biggest issues is that it doesn’t seem to have any memory, so every conversation starts from scratch.
Is there a way for a Custom GPT to have memory?
At the moment, no, a Custom GPT built with the GPT Builder does not have persistent memory across sessions. OpenAI’s Help Center says memory is “not currently supported in custom GPTs,” so each new chat with that GPT starts fresh, even if the user has Memory turned on in ChatGPT.
Hey @taskshrimp, as @polepole mentioned, memory isn’t currently supported in Custom GPTs. So even if memory is turned on in ChatGPT, conversations with a Custom GPT are still stateless.
If you're looking for continuity (for example with something like language learning), using Projects is probably the best workaround right now.