Chroma is the most developer-friendly vector database available — it runs in-process in Python with zero configuration, making it perfect for prototyping, Jupyter notebooks, and small-to-medium applications. Think of it as SQLite for vectors: simple, embedded, zero-ops, and surprisingly capable for its category.
Install via pip, import ChromaDB, create a collection, add documents. Chroma auto-generates embeddings using sentence-transformers by default (no API key needed) or accepts pre-computed vectors from any source. Persistence is handled automatically via DuckDB and Parquet files. For production, there's also a client-server mode where Chroma runs as a separate service and accepts API calls. Managed cloud version is in beta.
Completely free and open source. Self-hosted has no license fees. Hosted cloud (Chroma Cloud, in beta) will have a free tier and paid tiers — pricing TBD.
Thousands of developers for prototyping and small production deployments. Popular in research, Jupyter notebooks, and RAG tutorials. Integrated with LangChain and LlamaIndex by default.