LEANN open-source vector index cuts a 60M-chunk RAG index from 201 GB to 6 GB
AIUC Berkeley's Sky Computing Lab released LEANN, an open-source vector index that stores a pruned graph and recomputes embeddings at query time instead of saving them. The repo benchmark shows a 60 million chunk Wikipedia index shrinking from 201 GB to 6 GB, with retrieval taking seconds rather than milliseconds on the tested hardware.
Source details
MarkTechPost · Asif Razzaq · marktechpost.comPublished · added here
Original title: Meet LEANN: An Open-Source Vector Index that Shrinks a 60M-Chunk RAG Index from 201 GB to 6 GB