Milvus
used in
0
stacks
Open-source, cloud-native vector database for large-scale ANN search, with Zilliz Cloud as its managed offering.
Apache 2.0 licensed vector database governed under LF AI & Data, written primarily in Go and C++. Ships in three deployment modes: Milvus Lite (pip-installable, in-process), Standalone (single machine), and Distributed (horizontally scalable to billions of vectors). Supports metadata filtering, hybrid search, and multi-vector queries, with integrations for LangChain, LlamaIndex, and major embedding providers. Zilliz, the company behind Milvus, offers Zilliz Cloud with serverless, dedicated, and BYOC options including a free tier.
Strengths
- Scales to billions of vectors via a disaggregated, distributed architecture
- Three deployment modes span prototyping (Lite) to enterprise clusters
- Broad index-type support and hybrid/multi-vector search
- Neutral governance under LF AI & Data with a large contributor base
Tradeoffs
- Distributed mode depends on multiple external components (etcd, message queue, object storage), making self-hosted operations complex
- Heavyweight for small or single-node workloads where simpler stores suffice
- Feature differences between Lite, Standalone, and Distributed modes can complicate migration paths