VectorChord offers
- 1-but quantization search that speeds up search candidates and then search with full precision, that yeals significant overall retrieval latency improvement while not sacrificing accuracy.
- Storing the Quantized vectors in the DB significantly reduces storage while sacrificing some of the recall accuracy (not released yet, available in the main branch)
- BM25 Full Text Search available for HybridSearch (Vector + FTS)
https://github.com/tensorchord/VectorChord