← Templates · Databases
MI
Milvus
Community3 services
Open-source vector database built for AI — similarity search and storage for embeddings at scale.
| Runs | 3 services |
|---|---|
| Stack | milvus-etcdmilvus-miniomilvus-standalone |
| Provisioned for it | Nothing, it's self-contained |
| Verified | Not yet |
Why deploy on Railyard
Railyard provisions all 3 services in this stack together — milvus-etcd, milvus-minio, milvus-standalone — wired to reach each other over the private network, on a server you already pay for.
Deploying Milvus is one piece of a bigger stack you can keep building — add a database, a worker, a cache, or another app, all on the same servers you already pay for.
Good for
- Vector similarity search for embeddings at scale (RAG, recommendation, image/audio search)
- A self-hosted alternative to a managed vector database when you already run everything else on Railyard
- Storing and querying billions of vectors with metadata filtering
Good to know
- etcd (metadata) and MinIO (object storage for segments) are required dependencies, not optional — Milvus has no single-container mode that skips them, even in "standalone" deployment
- Give this stack's server real memory headroom — Milvus's own sizing guidance starts at 4GB+ of free RAM for a small standalone deployment and grows with collection size
- Connect with any Milvus SDK (pymilvus, etc.) at this stack's milvus-standalone service on port 19530 (not exposed by this template's public port, which is the 9091 health/metrics endpoint) — reach it from other apps on your team via its private domain