how to cut ConversationBufferMemory

[Zanejins]

The code I wrote is as follows. Since all the conversation records are saved to the memory, the conversation is too long for a while, and then an error is reported. Is there any way to make the memory only save the last few Q&A records to solve this problem?

self.llm = OpenAI(temperature=0, max_tokens=2048)
self.memory = ConversationBufferMemory()
self.conversation = ConversationChain(
            llm=self.llm,
            verbose=True,
            memory=self.memory,
        )

Error:

[Ludacon]

Ive had this error break otherwise very productive runs. AFTER restarting it also seems to get lost easier.

[sohrabsaran]

Is this issue still there in the latest codebase?

The code base will take time to understand but just sharing my ideas about what the code logic should be.

There will need to be a way to divide up the original problem into sub-problems.

And sub-problems may match different parts of the conversation.
The matched part of the conversation will need to be analyzed further to confirm its context and relevance to the current sub-problem.

There will also need to be a way to track and use the solved sub-problems, to solve the original parent problem.

There may even be a need to read only a part of a single dialog of the conversation, extract the needed information from it, and then note that down. And then continue to read the remainder of the dialog.

Now, for those who are familiar with the code, how much of these ideas are already present in the codebase?

May be we can look at file histories and ask the original code contributors to get a faster understanding of the code design and limitations.