vectordb

Why we included this project

vectordb is a Python library that gives you a working vector index for semantic search over embeddings without pulling together several moving parts yourself. It leans on DocArray for the nearest-neighbor search logic and Jina for serving, and covers the usual CRUD operations plus sharding and replication. You can run it in-process with an in-memory or HNSW index, or expose the same database over gRPC, HTTP, or WebSocket and talk to it from a thin client. That dual mode suits teams that want to prototype locally and later move the same code to a hosted or on-premise service. If you already work with dataclass-style document definitions, the typed DocArray schema keeps the API familiar and predictable.