Chroma
4.4
Chroma is an open-source search engine that handles vector, full-text and metadata search in one interface. It is the usual first choice for prototyping retrieval because it runs in-process with no infrastructure at all, and scales to a server when the prototype survives.
- Commercial use OK
- Free tier
- Pricing
- Free
- Free tier
- Free tier
- Commercial use
- Permitted
- Japanese UI
- Not confirmed
- Enterprise contract
- Not confirmed
What it is good for
- Prototyping RAG before committing to infrastructure
- Local development where a hosted database is overkill
- Applications small enough to embed the store in the process
- Teams that want vector and keyword search behind one API
Strengths and trade-offs
Strengths
- Runs in-process — nothing to provision to get started
- Vector, full-text and metadata search in one interface
- Open source with no licence cost
- Short path from prototype to a server deployment
Trade-offs
- Not aimed at very large corpora
- Operational tooling is thinner than the commercial options
- You own scaling and backup decisions
Editors' review
The one to start with, because starting costs nothing
Most retrieval projects die before they prove anything, because step one is provisioning a database. Chroma removes that: it runs inside the application, and the question becomes whether retrieval helps rather than whether the infrastructure is worth it.
Having vector, full-text and metadata search behind one interface matters more than it looks. Real queries mix all three, and discovering that after choosing a vectors-only store is an expensive detour.
It is not where large production corpora end up. Treat it as the way to find out what you actually need, then move to a managed store with that knowledge.
Verdict
The right first step for any retrieval project. Plan the migration once the corpus is large.
Compiled from official documentation and public sources, with our own editorial judgement. We do not publish figures we cannot source.