Milvus

By Zilliz

Open-source vector database built for billion-scale similarity search

Website ↗ Repository ↗ Docs ↗

v2.6.202026-07-14GoApache-2.0

#vector-db #embeddings #rag #go

45.4k Open source Free tier Paid

Doc version 1.0 · Updated June 28, 2026

Related tools

Overview

Milvus is an open-source vector database designed for similarity search at massive scale, with a distributed architecture that handles billions of vectors.
It runs embedded (Milvus Lite), self-hosted, or as managed Zilliz Cloud — the same API across all three.

flowchart LR
  vecs[vectors] --> m[(Milvus)]
  q[query vector] --> m
  m --> top[top-k]

The Code samples tab shows the embedded Milvus Lite flow.

When to use it

Choose Milvus when you expect very large collections and need a battle-tested distributed vector store — prototype with Milvus Lite, then scale to a cluster or Zilliz Cloud without changing your code.

Pricing

Open source (Apache-2.0), free to self-host. Managed Zilliz Cloud adds serverless and dedicated options.

Plan	Price	Notes
Free	$0	5GB, managed
Serverless	Pay-per-use	Scales to zero.
Dedicated	from $99/mo	Reserved clusters

Source ↗ · as of 2026-06-27

Example

Python · Milvus Lite

When to use

Use Milvus Lite to prototype locally with the same API you scale to a clustered Milvus or Zilliz Cloud deployment.

How it flows

Create a collection with a dimension, insert vectors with fields, then search by a query vector for the top-k.

Structure

flowchart LR
  vecs[vectors] --> m[(Milvus)]
  q[query vector] --> m
  m --> top[top-k]

from pymilvus import MilvusClient

# Milvus Lite runs embedded from a local file — no server to start.
client = MilvusClient("demo.db")
client.create_collection("docs", dimension=4)
client.insert("docs", [{"id": 1, "vector": [0.1, 0.2, 0.3, 0.4], "text": "hello"}])

hits = client.search("docs", data=[[0.1, 0.2, 0.3, 0.4]], limit=1, output_fields=["text"])
print(hits[0][0]["entity"]["text"])

Example

Milvus Lite — embedded vector search

A real, runnable mini-project. Download it and run with Docker.

View samples/milvus_1 on GitHub ↗

June 28, 2026

Milvus Lite — embedded vector search

A tiny Milvus script using Milvus Lite: it creates a collection, inserts a few vectors, runs a similarity search, and prints the closest row. Embedded from a local file — no server, no API key, no network — so the output is deterministic.

Run with Docker

cd samples/milvus_1
docker build -t aas-milvus .
docker run --rm aas-milvus

Run with Docker (in a devcontainer with DooD)

In a dev container that talks to the host Docker daemon (Docker-outside-of-Docker), the foreground docker run above often prints nothing and exits 0 — but the run itself succeeds. Run detached and follow the logs instead:

cd samples/milvus_1
docker build -t aas-milvus .
docker logs -f "$(docker run -d aas-milvus)"

Run locally

cd samples/milvus_1
pip install -r requirements.txt
python app.py

The same MilvusClient API scales from Milvus Lite to a self-hosted Milvus cluster or Zilliz Cloud — just pass a server URI instead of a file path.

Example run

Milvus Lite runs locally and this sample is deterministic, so the output is the same every run.

nearest: Cats are independent pets that purr.

Files

app.py

"""Milvus Lite: embedded Milvus from a local file — no server to start.

Creates a collection, inserts a few vectors, runs a similarity search, and
prints the closest row. Fully local and deterministic: no API key, no network.

Run:
    docker build -t aas-milvus .
    docker run --rm aas-milvus
"""

from pymilvus import MilvusClient


def main() -> None:
    client = MilvusClient("/tmp/milvus-demo.db")  # Milvus Lite — embedded file
    if client.has_collection("docs"):
        client.drop_collection("docs")
    client.create_collection("docs", dimension=4)
    client.insert(
        "docs",
        [
            {"id": 1, "vector": [0.1, 0.2, 0.3, 0.4], "text": "Cats are independent pets that purr."},
            {"id": 2, "vector": [0.9, 0.8, 0.7, 0.6], "text": "A rocket is propelled by ejecting exhaust."},
        ],
    )

    hits = client.search("docs", data=[[0.12, 0.21, 0.29, 0.41]], limit=1, output_fields=["text"])
    print("nearest:", hits[0][0]["entity"]["text"])


if __name__ == "__main__":
    main()

Dockerfile

# Minimal image for the Milvus Lite sample. Embedded — no services, no keys.
# Python 3.12 for the widest milvus-lite wheel availability.
FROM python:3.12-slim

ENV PYTHONUNBUFFERED=1

WORKDIR /app

COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

COPY app.py .

ENTRYPOINT ["python", "app.py"]