Practical overview
When it fits
Use Milvus when an AI application needs a dedicated vector-storage and retrieval layer, especially for collections that may require distributed or Kubernetes-native deployment. The documented Standalone and Milvus Lite paths also support single-machine or local experimentation.
Who it is for
- AI application developers
- Teams building semantic or image similarity search
- Engineers creating retrieval-augmented generation and knowledge systems
- Operators needing self-hosted or distributed vector-search infrastructure
Documented uses
- Store embeddings with associated scalar and JSON data.
- Run approximate nearest-neighbor vector searches.
- Filter or combine vector searches with metadata and hybrid search.
- Provide retrieval infrastructure for text, image, and multimodal AI applications.
Requirements & limits
The repository is public and identifies the project as Apache-2.0 licensed.
- Choose the local Milvus Lite path for a Python quickstart using a local database file.
- Use Standalone mode when the documented single-machine deployment model fits the workload.
- Consider the distributed Kubernetes-native architecture when independent scaling of query and data workloads is relevant.
- Metadata filtering and hybrid search are documented alongside vector search.
- CPU and GPU hardware acceleration are described as part of the search implementation.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Create a local Milvus Lite vector database and search a collection
Set up a local Python-backed database, create a collection, insert data, and run a vector search.
Before you start
- Python environment with the pymilvus package installed.
- Vector data whose dimension matches the collection configuration.
- An embedding function or query vectors for the search step.
Steps
Install the Python SDK with the command below. For a local Milvus Lite database, install
pymilvus[milvus-lite].pip install -U pymilvusInstantiate
MilvusClientwith a local filename such asmilvus_demo.db.Create a collection with
client.create_collection, specifying its name and vector dimension.Insert records with
client.insert.Encode query text into vectors and call
client.search, specifying the collection, query vectors, result limit, and output fields.
Expected outcome
A local collection contains inserted vector data and returns ranked search results for supplied query vectors.
Guide limitations
- The documented example assumes query vectors with a stated 768-dimensional example; collection dimensions must match the vectors used.
- The guide covers the local quickstart path, not production deployment or workload sizing.
Compare options for Vector search
These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.
| Project | Documented purpose | Platform | Access and requirements |
|---|---|---|---|
| Milvus | Store embeddings with associated scalar and JSON data. | Cloud, Docker, Self-hosted, Server | The repository is public and identifies the project as Apache-2.0 licensed. Documentation covers self-hosted, Docker, Standalone, Milvus Lite, and managed-service connection paths; the supplied evidence does not establish pricing or account requirements. |
| Mettis | Index text documents through an HTTP endpoint. | Linux, Self-hosted, Server |
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 30 September 2026
- Original website title
- milvus-io/milvus · Official homepage title · retained in its source language