I wrote a hybrid vector store from scratch in Go. Supports hybrid retrieval over BM25, Flat, HNSW, IVF, PQ and IVFPQ Index with Quantization, Pre-Filtering, Reranking, Reciprocal Rank Fusion, Soft Deletes, Index Rebuilds and much much more.
It's called Comet. Built for hackers, not hyperscalers. It's tiny enough to fit in your head, and good enough to blow it :D
It packs everything you need to understand how modern vector databases work internally, from indexing to rank fusion, querying, and even persistence.
https://github.com/wizenheimer/comet
https://pkg.go.dev/github.com/wizenheimer/comet
Star it, fork it, break it, fix it — make it your own
It's called Comet. Built for hackers, not hyperscalers. It's tiny enough to fit in your head, and good enough to blow it :D
It packs everything you need to understand how modern vector databases work internally, from indexing to rank fusion, querying, and even persistence.
https://github.com/wizenheimer/comet
https://pkg.go.dev/github.com/wizenheimer/comet
Star it, fork it, break it, fix it — make it your own