Chroma
Chroma is an open-source embedding database built for retrieval-augmented generation, storing vector embeddings alongside metadata and documents so applications can run similarity search and filtering over them. It offers both an embedded in-process mode and a client-server deployment.
Baseline topologies
| Topology | Minimum nodes | Roles |
|---|---|---|
| local-or-embedded | 1 | chroma |
| shared-retrieval-server | 1 | chroma |
What drives the cost
- Vector count (millions)
- Approximate number of stored vectors before growth overhead.
- Embedding dimension
- Dimension of the embeddings stored in the primary collections.
- Peak query concurrency
- Expected peak concurrent retrieval requests against the Chroma deployment.
Where it runs
bare_metal, dedicated, cloud_vm
Size Chroma and get a cost estimate