Chroma
Free tierOpen-source search infrastructure for AI — vector, full-text, regex, and metadata search at scale
Free tier available·All audiences·API available·Open source
Key strengths
Multi-modal search: vector, full-text (BM25/SPLADE), regex, and metadata in one systemBuilt on object storage (S3/GCS) — up to 10x cheaper than memory-based alternativesServerless and zero-ops — auto-scales with no manual tuning requiredApache 2.0 open-source with 27k GitHub stars and 15M+ monthly downloadsEnterprise-grade: SOC 2 Type II, BYOC, multi-region replication, customer-managed encryption keys
Free tier + paid plans
Self-hostable
No ratings yet
Technical Documentation
Installation
# Python
pip install chromadb
# JavaScript / TypeScript
npm install chromadb
Core Concepts
- Collections: Named groups of records (up to 5M records each, 1M collections per database)
- Embeddings: Dense or sparse vectors stored alongside documents and metadata
- Search Types: Vector (KNN), sparse (BM25/SPLADE), full-text (trigram/regex), metadata filtering
Search API Example (TypeScript — Sparse BM25)
const sparseRank = Knn({ query: "ML", key: "sparse_embedding" });
const search = new Search().rank(sparseRank).limit(10).select(K.DOCUMENT, K.SCORE);
const results = await collection.search(search);
Performance Specs
| Metric | Value |
|---|---|
| p50 latency (warm, 100k vectors) | 20ms |
| p99 latency (warm) | 57ms |
| Write throughput (per collection) | 30 MB/s (2000+ QPS) |
| Concurrent reads (per collection) | 10 (200+ QPS) |
| Recall | 90–100% |
Infrastructure Architecture
- Query layer: In-memory cache (hot) → SSD cache (warm) → S3/GCS (cold)
- Storage: All vectors, metadata, and indexes on object storage
- Tiering: Query-aware automatic data tiering and caching
- Deployment options: Chroma Cloud (managed), self-hosted OSS, BYOC Enterprise (your VPC)
Enterprise Features
- BYOC (Bring Your Own Cloud) deployment in your VPC
- Multi-cloud / multi-region replication
- Point-in-time recovery
- AWS PrivateLink support
- Customer-managed encryption keys (CMEK)
- SOC 2 Type II certified