LocalAI · Architecture

How LocalAI works

ONE API

many engines

Clients speak one API. The core routes each request. Every backend is a separate process, pulled only when a model needs it.

localai.io/docs/overview