#hnsw

Databases Bug Report

pgvector HNSW with a WHERE filter returns fewer rows than LIMIT

Ryan Pham · Mar 12, 2025 · 6 min read

Small tenants in our RAG product got 2 or 3 chunks back instead of 10, and sometimes none. HNSW finds ef_search candidates first and applies the tenant filter afterwards. Iterative index scans in pgvector 0.8 plus partial indexes for big tenants fixed recall.

Databases Bug Report

pgvector similarity search ignores the HNSW index: operator and opclass mismatch

Ryan Pham · Feb 5, 2025 · 6 min read

Semantic search in our docs assistant took 1.9 seconds and returned odd rankings. The query sorted by L2 distance while the HNSW index was built for cosine, so Postgres scanned every row. Matching the operator to the opclass brought it to 14 ms.