Pg_vectorsearch: An index that is faster than pgvector's IVFFlat and HNSW

Automated excerpt

RaBitQ two-stage filtering: estimate → error bound → rerank Install the extension first; see Building. Store vectors in pg_vectorsearch's own vec32 (float32) or vec16 (float16) columns. Raise for recall, lower for speed. table, with the matching vec16_* operator classes. columns, you can build a PRISM index on it directly.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.