Milvus

Milvus is an Apache-2.0, cloud-native vector database (Go) designed for large-scale approximate nearest-neighbor search, with storage/compute separation, sharding, and GPU index support. Milvus Lite — bundled inside pymilvus — embeds the engine in a local file for zero-infrastructure prototyping, while the full standalone/clustered deployment targets production scale.

It is maintained by Zilliz with one of the largest open-source vector-search communities, broad SDK support (Python, Java, Go, Node, C#), and documented integrations across LangChain, LlamaIndex, and Semantic Kernel.

1. Install the client (bundles Milvus Lite, no server needed): pip install -U pymilvus (Python 3.8+)
2. Create a local file-backed database: from pymilvus import MilvusClient; client = MilvusClient("milvus_demo.db")
3. Create a collection: client.create_collection(collection_name="demo_collection", dimension=768) (default metric is COSINE)
4. Insert entities: client.insert(collection_name="demo_collection", data=[{"id": 0, "vector": [...], "text": "...", "subject": "history"}])
5. Search with metadata filtering: client.search(collection_name="demo_collection", data=[query_vector], limit=2, output_fields=["text", "subject"], filter="subject == 'history'")
6. For production, run the standalone Docker stack with the official compose file: wget https://raw.githubusercontent.com/milvus-io/milvus/master/deployments/docker/standalone/docker-compose.yml && docker compose up -d