GigaVector
A vector database library written in C with Python bindings, covering twelve indexing algorithms, five distance metrics, and search features such as hybrid, filtered, geo-spatial and grouped search.
GigaVector is a vector database library written in C, with Python bindings, covering indexing, search, storage, graph, networking, security and AI integration.
It offers twelve index types, from exact KD-Tree search to IVF-PQ, IVF-Flat, IVF-SQ8, IVF-TurboQuant, IVFDisk, DiskANN, PQ, LSH and sparse indexes, each suited to different dataset sizes and memory budgets, plus an automatic index-suggestion helper. It supports five SIMD-optimized distance metrics (Euclidean, Cosine, Dot Product, Manhattan, Hamming) and search features including k-NN, range, batch, filtered, hybrid vector-plus-BM25, grouped and geo-spatial search, late-interaction ColBERT-style scoring, recommendation queries, MMR diversity reranking, custom ranking expressions with decay functions, and a SQL query interface.