Documentation
minibrain trains, evaluates and serves models of any size on your own data, on your own hardware or cloud. Start with the install, then follow a model from data to a live endpoint.
InstallOne command, pip, uv or Docker, on Linux or macOS.
ModelsBase models, where they come from and how they are kept.
ServingEngines, pools, the gateway and the OpenAI-compatible API.
SecurityAccounts, single sign-on, secrets, audit and signed releases.
ObservabilityMetrics, traces and logs into your own stack.
ArchitectureHow main, the gateway, workers and engines fit together.