VexDB
A vector similarity search engine that plugs into PostgreSQL, DuckDB, and SQLite, sharing the same graph index algorithm, SIMD distance calculations, and quantization across all three backends.
VexDB-Lite adds vector similarity search to PostgreSQL, DuckDB, and SQLite as an extension for each. The PostgreSQL extension adds a floatvector type, distance functions and operators, a graph-based index, and options such as parallel index building and PQ or RaBitQ quantization, with an optimizer rewrite into an ANN index scan. The DuckDB extension adds a similar GRAPH_INDEX on float array columns with metadata filtering, and a SQLite build is also provided.
All three backends share the same underlying graph index algorithm, SIMD-dispatched distance functions, and product- and RaBitQ-quantization kernels, so behavior and tuning options are consistent across the database it is embedded in.
It is written in C++ and released under the MIT license.