ChromaDB — векторная база данных для RAG

★ 14k

ChromaDB is an open-source vector database and the simplest way to start putting embeddings somewhere and searching them by meaning today. It suits prototypes, local assistants and small RAG pipelines: install it with one command, run it in-process with no separate server, and keep the data in a file on disk. The whole API rests on four operations — add documents, update, delete, and query a collection for similar items. Chroma can handle vectorisation for you: hand it text and it computes embeddings with a bundled model, so your first working search needs no external service, though you can plug in your own model. A client-server mode is available when the app and the store should live apart. Choose it for speed of setup and simplicity; once a collection grows into tens of millions of vectors and you need high availability, teams usually migrate to a heavier database.