vectordb

Why we included this project

VectorDB is a small Python library that does embeddings-based text retrieval entirely in-process. You hand a Memory object some text, it chunks long passages, computes embeddings locally, and runs similarity search, returning the closest chunks with whatever metadata you attached. There's no separate database service to run, which makes it a good fit for a script or small service that needs semantic search over a few thousand documents. You can pick an embedding model by quality tier or name any HuggingFace model, and point it at a file if you want the index to survive restarts. It was built to power AI features inside Kagi Search, so the API stays deliberately small and lookups are fast; that also makes it a reasonable place to start when you're prototyping retrieval and aren't sure yet whether you need a heavier vector store.