milvus — Milvus is a high-performance, cloud-native vector database designed for scalable approximate nearest neighbor (ANN) search, powering AI applications with unstructured data.

Milvus is a high-performance, cloud-native vector database designed for scalable approximate nearest neighbor (ANN) search, powering AI applications with unstructured data.

About this project

Milvus is a distributed vector database written in Go and C++ that enables efficient organization and searching of vast amounts of unstructured data, including text, images, and multi-modal information. It is designed for high scalability and availability, utilizing a K8s-native architecture that separates compute and storage to allow independent horizontal scaling. Key capabilities include: - Vector Search & Indexing: Supports multiple index types such as HNSW, IVF, FLAT, SCANN, and DiskANN, with hardware acceleration for both CPU and GPU (including NVIDIA CAGRA). - Hybrid Search: Natively supports both dense vectors for semantic search and sparse vectors for full-text search (BM25, SPLADE, BGE-M3), allowing for reranked hybrid results. - Data Management: Features multi-tenancy isolation, hot/cold storage mechanisms for cost-effectiveness, and support for scalar data types (integers, strings, JSON) for metadata filtering. - Security: Implements mandatory user authentication, TLS encryption, and Role-Based Access Control (RBAC). - Deployment Options: Available as a distributed cluster, a standalone Docker deployment, or a lightweight Python version (Milvus Lite). Milvus integrates with popular AI frameworks like LangChain, LlamaIndex, OpenAI, and HuggingFace, and provides a Python SDK (pymilvus) for data ingestion and querying.