Vector search uses vector embeddings to search for similarities. Although vector embedding is an array of numbers, traditional search algorithms that handle exact matches are not suitable for searching for similarities. Vector search algorithms such as Hierarchical Navigable Small World (HNSW) and Inverted File (IVF) allow approximate matches , and the availability of such algorithms in databases is essential for utilizing vector search. FerretDB supports vector search by enabling vector index creation using HNSW and IVF algorithms. In this talk, I will describe how HNSW and IVF algorithms work to help selection of algorithms suitable for applications, and demonstrate how to configure and use vector search in FerretDB.
