Milvus
ActiveDescription
Milvus is a high-performance open-source vector database built for AI applications. It supports storage, indexing, and similarity search of large-scale vector data, ideal for RAG, recommendation systems, and more.
Key Features
- High-performance Vector Search — HNSW, IVF, FLAT, DiskANN index types with CPU/GPU hardware acceleration
- Distributed Architecture — Compute-storage separation, K8s-native, horizontal scaling with high availability
- Multi-tenancy Support — Isolation at database, collection, or partition level, supporting millions of tenants
- Hybrid Search — Vector search combined with scalar filtering, metadata filtering and range search
- Milvus Lite — Lightweight Python library, quick local startup via pip install
- Real-time Streaming Updates — Real-time data insertion and index updates keeping data fresh
Use Cases
💡 Build large-scale RAG systems providing external knowledge retrieval for LLMs
💡 Implement similar item retrieval and personalized recommendations in recommendation systems
💡 Build multimodal search systems supporting similarity search for text, images, and unstructured data
💡 Deploy enterprise semantic search engines with real-time queries on billions of vectors
💡 Set up vector storage infrastructure for AI applications managing embeddings and metadata
Strengths & Limitations
✅ Strengths
- • Actively maintained, recent updates
- • High community interest (45.7k stars)
- • Permissive open-source license (Apache-2.0)
Categories
Quick Start
# Install pymilvus
pip install -U pymilvus
# Create local vector database with Milvus Lite
from pymilvus import MilvusClient
client = MilvusClient("milvus_demo.db")
# Create a collection
client.create_collection(
collection_name="demo",
dimension=768,
)
# Insert data
client.insert(collection_name="demo", data=data)
# Perform vector search
res = client.search(
collection_name="demo",
data=query_vectors,
limit=5,
)