Chroma

Chroma is an open-source database for AI applications that combines vector, full-text, regex and metadata search. Developers use it to store documents and embeddings and retrieve the relevant ones for LLM apps and agents. In October 2026 the Chroma team announced it is joining Replit: the open-source project stays Apache 2.0 and Replit will keep contributing to it, while the hosted Chroma Cloud service is being wound down.

Chroma's core API is four functions: create a collection, add documents, then query or get them. It handles tokenization, embedding and indexing automatically, or takes embeddings you bring yourself. It runs in memory inside a Python or JavaScript app for prototyping, with persistence on disk, or as a client-server setup with chroma run. Search covers dense and sparse vectors (BM25, SPLADE), full-text and regex matching, and metadata filtering. The core is written in Rust.

Chroma is the easiest vector database to start with: one pip install and four functions, and it runs inside your app before you ever need a server. It is Apache-licensed, and the open-source version continues after the Replit deal.

Stars

0

Forks

0

Commits

0

Repository

chroma-core/chroma

• PineconeManaged vector database for AI applications.
• MilvusNo. 215Open-source vector database built for scale.
• QdrantNo. 150Open-source vector database written in Rust.
• WeaviateNo. 033Open-source AI-native vector database.

See every open-source alternative to Pinecone in the collection.