Milvus

Milvus is an open-source vector database purpose-built for massive-scale similarity search, managing billions of vectors across distributed clusters with the throughput and reliability characteristics required for production AI systems at enterprise scale. Where tools like Chroma and Qdrant optimize for developer simplicity and moderate scale, Milvus is architected for horizontal scaling: its disaggregated storage and compute model allows query nodes, data nodes, and index nodes to scale independently based on workload characteristics. Milvus handles multiple vector types (dense, sparse, binary), multiple index algorithms (HNSW, IVF, DiskANN), and hybrid scalar-vector queries in a distributed system that maintains high availability through its Kubernetes-native design.

The Milvus Lite version runs in-process for local development, the standalone version handles single-server deployments, and the distributed version scales to petabyte-scale production systems, three deployment modes covering the full range from development through enterprise scale. Zilliz Cloud provides a fully managed Milvus service for organizations that want Milvus's capabilities without operating distributed infrastructure. The Apache 2.0 license permits commercial use without restrictions.

Milvus is most commonly deployed in large-scale recommendation systems, enterprise search applications, image and video similarity search, and RAG systems serving millions of daily queries where horizontal scalability and consistent sub-100ms latency at billion-vector scale are requirements.