# Caliby

An embeddable vector database with document storage, hybrid search, and metadata filtering, designed to scale beyond available memory without running as a separate server.

Caliby is a vector database library that runs inside an application rather than as a client-server system, aimed at cases where data grows larger than can fit in RAM. It combines document storage, vector search, and filtered search, and is designed to degrade smoothly as data exceeds memory rather than failing or slowing sharply once it does.

It installs with pip and needs no separate infrastructure such as Docker or a running server, so it can be bundled directly into desktop apps, edge devices, or microservices, and is intended for uses such as persistent AI agent memory, local-first semantic search in desktop or mobile apps, code search and documentation retrieval inside IDEs and CLIs, and edge deployments without network access.

It is written in C++ and released under the MIT license.

Website: https://github.com/zxjcarrot/caliby