Glossary term
Glossary term
Memory and Retrieval
Hierarchical Navigable Small World - graph-based approximate nearest neighbour index providing logarithmic search complexity.
Qdrant uses HNSW as its default index for vector search - achieving 99% recall at <5ms query latency for 1M-vector indexes, used by Zendesk for knowledge-base semantic search across support ticket history.
Weaviate's HNSW implementation supports dynamic updates (insert/delete) without rebuilding the entire index - used by news organisations to index breaking stories in real time as they are published.
pgvector 0.7+ uses HNSW indexing as its primary approximate search algorithm - enabling PostgreSQL users to serve 10M+ vector search queries per day without a separate vector database.