Milvus

platform active open-source

An open-source vector database built for scalable similarity search and AI applications. Milvus supports billion-scale vector search with high performance, offering both standalone and distributed deployment modes. It provides multiple index types (IVF, HNSW, DiskANN) and supports hybrid search combining vectors with scalar filtering.

Implements

Concepts this tool claims to implement:

  • Core vector similarity search with support for L2, IP, COSINE metrics. Multiple index types including IVF_FLAT, IVF_SQ8, IVF_PQ, HNSW, DiskANN. Billion-scale vector search with distributed architecture.

  • Supports combining vector similarity with scalar attribute filtering. Boolean expressions for metadata filtering during search.

  • Embedding secondary

    Stores and indexes embedding vectors. Does not generate embeddings directly but integrates with embedding providers.

Integration Surfaces

  • Python SDK (pymilvus)
  • Java SDK
  • Go SDK
  • Node.js SDK
  • REST API
  • gRPC API

Details

Vendor
Zilliz
License
Apache-2.0
Runs On
local, cloud, hybrid
Used By
human, agent, system

Notes

Milvus is designed for production-scale vector search workloads. Zilliz Cloud offers a fully managed version. The standalone mode is suitable for smaller deployments while the distributed mode handles billions of vectors.