Find answers from the community

Updated 2 months ago

I was just playing with RAPTOR but I'm

I was just playing with RAPTOR but I'm having trouble conceptualizing how RAPTOR based indexes could be updated long-term. Essentially, whenever you have new material, you would have to start summarizing the new material and on each round pull existing summarizations based on similarity and then do another round to summarize those summarizations. This will get very expensive and won't scale. Am I thinking this wrong?
L
m
3 comments
Yea they aren't really made to be updated

Updating would require re-clustering the entire dataset imo
Its a neat POC/idea, not sure how practical it is
Yeah, I have to agree. I can't really think of many production use cases for it tbh.
Add a reply
Sign up and join the conversation on Discord