Your own AI, on your own hardware
Ready to work for you. Install LocalAI, run a first model, then give an agent your files. Pages that need an add-on or an administrator switch say so.
Start in three steps
1
Install
Use the script on Linux, or on a Mac with Apple Silicon. Other ways: a container, the macOS app, Kubernetes. See the install guide.
2
Run a first model
LocalAI downloads the model and picks the engine for your machine. See the quickstart.
3
Build an agent
Open the web interface on port 8080. Go to Agents, choose New agent, then give it instructions. See Build your first agent.
What you can do
Built in is part of LocalAI. Add-on comes from the add-on library, and an administrator installs it. Admin on is built in, but on a server with accounts an administrator must switch it on. With authentication off, you are the administrator.
The add-ons are in the development channel of the add-on library in the current development builds.
Go further
Find your way
- Start here: install, first model, first agent, API examples.
- Concepts: architecture, users and sharing.
- Features: what you can do, grouped by task.
- Add-ons: install, use and build add-ons, and the catalogue of what each one adds.
- Models and backends: the gallery, what models can do, model configuration, backends and GPUs.
- Administration: authentication, settings, monitoring and troubleshooting.
- Reference: command line, binaries and the FAQ.
- Contributing: build from source and write docs.