"
Overview
Chroma is an open-source search infrastructure purpose-built for AI applications. It provides fast, serverless, and scalable search capabilities that support vector, full-text, regex, and metadata search — all built on top of object storage. With 15M+ monthly downloads, 27k GitHub stars, and an Apache 2.0 license, Chroma has become a trusted foundation for millions of developers building AI-powered products.
What sets Chroma apart is its zero-ops architecture. Unlike legacy search systems that require manual tuning and dedicated infrastructure, Chroma automatically scales with your data and traffic, offering intelligent data tiering between memory, SSD, and object storage. This means you get low-latency queries over billions of multi-tenant indexes without the engineering overhead of managing clusters.
Developers love Chroma for its simplicity and performance. With SDKs for TypeScript, Python, and Rust, plus a powerful CLI, integrating Chroma into your AI stack is straightforward. The platform delivers p50 query latencies as low as 20ms for warm queries, making it ideal for real-time AI applications.
Key Features
- Multi-Modal Search: Combines sparse vector search (BM25, SPLADE), dense vector search, full-text search (trigram and regex), and metadata filtering in a single unified query interface.
- Serverless Architecture: Built on object storage (S3/GCS) with automatic data tiering — no manual tuning, auto-scales with usage, and offers serverless pricing up to 10x cheaper than traditional search.
- High Performance: Handles billions of multi-tenant indexes with p50 latency of 20ms (warm) and supports 30 MB/s write throughput per collection with 90-100% recall.
- Dataset Versioning: Built-in forking capabilities for dataset versioning, A/B testing, and safe roll-outs of search configurations.
- Multi-Language SDKs: First-class support for TypeScript, Python, and Rust with intuitive APIs for adding documents and executing complex ranked searches.
