What is the best way to pass the conversation history to gpt-index. Is there a way to pass through something like ConversationalBufferWindowMemory to underlaying LLMChain?
not sure I've fully understood the request: is the idea to have tooling that enables a conversation buffer to itself be indexed? I think that's a potentially fun idea that would enable long conversations over multiple sessions. The prompt could include 50% recent conversation history, and 50% queries from a vector store over previous conversations. I'd imagine that'd e.g. enable an agent to remember your age from a previous conversation