Algorithm · Established · Advanced

HNSW indexing

A graph-based approximate nearest-neighbour index that gives vector search its speed.

What HNSW indexing is

Hierarchical Navigable Small World graphs connect vectors so search can hop quickly towards the nearest neighbours instead of scanning everything.

How it works

A layered proximity graph is built at insert time; queries descend layers greedily. Parameters trade index build time and memory against recall and speed.

Why it matters

It is the default index in most vector databases, and its parameters are the main tuning surface for retrieval quality.

Common uses

→Vector database indexing
→Real-time similarity search

Strengths

✓Fast with high recall
✓Supports incremental inserts

Watch for

✓Memory-heavy
✓Deletes are awkward

Continue exploring

Vector Database Embeddings Semantic Search