Milvus Vector Database

Milvus is an open-source vector database designed for scalable similarity search and AI applications. Created by Zilliz in 2019 and now a CNCF Graduated project, Milvus has over 30,000 stars as of 2026 and is used by over 1,000 organizations including leading internet companies. Milvus powers semantic search, image search, recommendation systems, chatbots, and drug discovery by storing and querying high-dimensional vectors from embeddings generated by deep learning models. Key features include: vector indexing (multiple index types including IVF_FLAT, IVF_SQ8, IVF_PQ, HNSW, DiskANN, ANNOY, FLAT, and GPU-aware indexes for different speed-accuracy tradeoffs), hybrid search (combining vector similarity search with scalar filtering, enabling queries like finding vectors within a range of metadata values), multi-vector search (multiple vector fields per collection with different similarity metrics in a single query), partition and partition key (data partitioning for efficient retrieval and multi-tenancy), dynamic schema (adding and removing fields without schema migration), upsert and delete (real-time data modification with consistency guarantees), consistency levels (Strong, Bounded Staleness, Session, and Eventually for different latency-consistency tradeoffs), distributed architecture (separation of storage and compute using S3-compatible object storage and message queues for horizontal scalability), cloud-native deployment (Kubernetes-native with Helm charts, Terraform modules, and managed cloud service Zilliz Cloud), SDKs (Python, Java, Go, Node.js, C++, Rust, C#, and RESTful API), and ecosystem integration (LangChain, LlamaIndex, Hugging Face, OpenAI, and Sentence Transformers for LLM-based retrieval pipelines).