← Back to Glossary
ANN (Approximate Nearest Neighbor)
Phase 03 · RAGtechnique

Algorithms that trade off a tiny fraction of accuracy to search millions of vectors 100x faster than brute-force comparisons.

Why it exists

Brute-force k-NN compares a query against millions of vectors one-by-one, which is too slow at scale.

How it works

Organizes vectors into spatial graphs (HNSW) or inverted clusters (IVF) to navigate directly to the most relevant region in milliseconds.

Where you'll use this