chroma

Why we included this project

Chroma is the database layer many teams reach for when they need to store and search embeddings without standing up a heavyweight infrastructure stack. The core API is deliberately small, just a few functions for creating a collection, adding documents, and querying for the most similar results, so a semantic search or retrieval feature is easy to prototype in an afternoon. By default it handles the embedding and indexing for you, including tokenization, and you can still pass your own embeddings when you have a specific model in mind. It ships as a pip-installable Python client with an in-memory mode for experiments and a client-server deployment for production, so the same code can grow from a notebook into a shared service. Teams building RAG pipelines, agent memory, or document search will find the metadata filtering and simple query API useful for getting something working quickly and then hardening it later.