Milvus — масштабируемая векторная база данных

★ 29k

Milvus is an open-source vector database written in Go and C++, built for the case where you have a great many embeddings and more keep arriving. Reach for it when semantic search has to run across hundreds of millions or billions of records: similar texts, images, audio and multimodal content, product-catalogue search, or a company-wide RAG store. The architecture is distributed and Kubernetes-native — storage, indexing and query run as separate nodes, so you scale by adding machines rather than migrating to a different database. It offers several index types trading speed against recall and memory, metadata filtering, and hybrid search. A lightweight mode runs without a cluster for smaller projects. Milvus does not compute embeddings itself; you plug in a model separately. Choose it when the workload will clearly outgrow a single machine — for a prototype or a modest collection, something simpler to operate is the better call.