Indexing sounds promising. I've been trying to work out a way to provide long term memory to an LLM because starting over fresh with each session is a pain and prompt chaining only goes so far. How does this indexing work / can you point me at a primer?
