Milvus
By Zilliz
Milvus is an open-source vector database designed for storing and searching large-scale embedding vectors, built to support similarity search and retrieval workloads such as semantic search and retrieval-augmented generation.
Definition
Milvus is an open-source vector database designed for storing and searching large-scale embedding vectors, built to support similarity search and retrieval workloads such as semantic search and retrieval-augmented generation.
Overview
Milvus was created to handle vector similarity search at large scale and is distributed as open-source software, hosted under the LF AI & Data Foundation, with commercial support and a managed cloud version (Zilliz Cloud) offered by its founding company, Zilliz. Because it's open-source, teams can self-host Milvus to keep full control over data and infrastructure, unlike fully managed-only services. Milvus supports multiple indexing algorithms for approximate nearest-neighbor search, letting users tune the tradeoff between search speed, accuracy, and memory usage depending on their dataset size and latency requirements. It integrates with common embedding and orchestration frameworks used to build RAG applications. As with other vector databases such as Pinecone and Qdrant, Milvus is typically used alongside an embedding model and an LLM to build semantic search or retrieval-augmented generation systems, a pattern covered in the Retrieval-Augmented Generation course.
Key Features
- Open-source vector database for large-scale similarity search
- Multiple ANN (approximate nearest-neighbor) indexing algorithms
- Self-hostable, with a managed cloud option (Zilliz Cloud) also available
- Horizontal scalability designed for billions of vectors
- Support for hybrid search combining vector and scalar filtering
- Integrations with common embedding models and AI frameworks
- Governed under the LF AI & Data Foundation as an open-source project