ann-search
Embeddable, in-memory, document-oriented database with a high-level Query builder interface.
Decides which partition a vector belongs in when a vector index is split across tenants, users, or time windows: three routing strategies plus result merge-back, one runtime dependency (pydantic), no index of its own. 243 tests, 98% coverage.
Embeded Vector Database for low performance devices
Lightweight approximate nearest neighbor (ANN) vector search engine and embedding indexing pipeline.
