FAISS Facebook AI Similarity Search
A library for fast similarity search across large collections of vectors, released by Meta in 2017. The engine under the hood of a large share of today's vector databases.
Exact search across billions of vectors is impossible in real time, so FAISS offers approximate indexes that trade a little accuracy for enormous speed. IVF partitions the space into clusters and searches only the nearest ones; HNSW builds a navigable multi-layer graph; product quantization compresses the vectors so they fit in memory, often to an eighth of their size. Indexes can be combined and run on GPUs. It is a library, not a database — no persistence, no in-place updates, no metadata filtering and no server component. If you need those you build around it, or use something that already has.