Anyone have good strategies for storing/accessing long term chat memory? I want to bring back only bits of conversations that are relevant to the topic in current conversation but am finding that a similarity search will bring back all kinds of extraneous and repetitive data. It’s all similar (so the vector search is definitely working), but it’s a needle in the haystack in terms of relevance.
I was thinking i should store only bits that are “significant”… but determining significance is no small feat.
Anyone have tips?