Vai al contenuto
AI.info

tools

Milvus

Database vettoriale open source per la ricerca per similarità, il recupero ibrido, il filtraggio dei metadati e i dati delle applicazioni di IA.

Milvus

In inglese

Milvus stores embeddings and structured data in collections, then supports vector, sparse, hybrid, text, and filtered search. It offers Lite for local Python applications, Standalone in Docker, and Distributed deployments on Kubernetes.

Developers use it for RAG, semantic search, image and multimodal search, deduplication, and AI-agent data stores. The software is free under Apache 2.0; managed hosting is offered separately through Zilliz Cloud and is not priced on Milvus.io.

Features

  • Dense, sparse, hybrid, and multi-vector search
  • Metadata filtering, scalar queries, and flexible schemas
  • TEXT fields with text_match, phrase_match, and BM25
  • External Collections for zero-copy lakehouse retrieval
  • Point-in-time read-only Snapshots
  • Server-side sorting, aggregation, and faceted search
  • REST and gRPC APIs with Python, Java, Go, C#, and Node.js clients
  • Apache 2.0 license

Use cases

  • Build retrieval-augmented generation applications
  • Search documents, images, audio, or other embedded data
  • Run semantic and hybrid search over product catalogs
  • Detect near-duplicate records with MinHash search
  • Prototype locally with Milvus Lite
  • Serve large-scale vector search on Kubernetes

Pros

    Cons

      Latest updates

      • milvus-3.0.4 (v3.0.4)

        Speeds up bulk import, compaction, and filtered and grouped search; adds lazy column-group loading for Storage V3 and fixes hangs and access control issues.

      • milvus-2.6.26 (v2.6.26)

        Speeds up RPC message decoding, reduces Proxy memory use for large search and query requests, and fixes crashes, hangs, and timestamp issues.

      • milvus-3.0.5 (v3.0.5)

        Adds resource-group-aware load status in QueryCoord, reduces memory use, and fixes stability issues around coordinator restarts, rolling upgrades, and etcd errors.

      • pkg/v3.0.5: fix: [3.0] Align milvus-common with Cardinal stream options (#54734) (v3.0.5)

        Aligned milvus-common with Cardinal stream options to enable native VectorLake image compilation.

      • milvus-3.0.3 (v3.0.3)

        Added Function Chain to Hybrid Search, element-level partial updates for Array, Struct, and JSON fields, and IS NULL filtering on nullable vector fields.

      Capabilities

      • Command line — “Milvus CLI” source
      • Self-hosted — “Milvus Distributed can be deployed on Kubernetes clusters, featuring a cloud-native architecture designed for billion-scale or even larger scenarios.” source
      • API — “RESTful API (official)” source
      • Official SDKs — “PyMilvus (Python SDK) (official)” source
      • Search over your data — “These vectors are then stored in a vector database, enabling fast and scalable searches and analytics.” source

      Get it

      Pricing

      Starting price
      Free
      Prices checked
      2026-10-25

      Milvus (Open Source)

      Free

      • 100% free open-source project under Apache License 2.0
      • Self-hosted deployments (Lite, Standalone, Distributed)
      • No usage limits from the software itself (hardware-dependent)
      Official website