Motrix Deploy

motrix-deploy runs exported policies without a training framework. It provides deployment artifacts, robot/task contracts, shared control execution, runtime plugin discovery, and the implementation behind mxlab deploy. Install motrixlab for the sole mxlab CLI; this component retains its Python APIs and internal callable deployment adapter. Use mxlab --version to query the version. Install the task and runtime plugins needed for your target. motrix-deploy does not depend on motrix-rl: training-run export belongs to motrix-rl, which depends on this package, while deployment runs independently. Create artifacts with mxlab export-deploy run_dir=<run-dir> output=artifacts/policy.deploy; see the deployment tutorial for export. Environment packages provide profile compilers through package registration callbacks in motrix_deploy.profiles; see adding a profile compiler for the provider recipe and discovery requirements.

With the Go2 task and a simulation backend installed:

mxlab deploy task=go2-walk-flat/sim artifact=artifacts/go2-walk-flat.deploy

Hardware API

Hardware backends return a fully constructed HardwareRuntime, independent of the robot model or transport:

from motrix_deploy.runtime.hardware import HardwareRuntime

runtime = HardwareRuntime(robot)  # An already constructed RobotInterface.

Construction does not open the robot connection. Bind a ControlSession before running; robot startup, enable, safe stop, and transport cleanup remain owned by the interface. The runtime uses the robot's optional gamepad device unless one is explicitly supplied.

For CLI recipes, hardware backend entry points in motrix_deploy.backends implement factory(config, context) -> HardwareRuntime. The context supplies the artifact's robot and control specifications. create_hardware_runtime validates and returns the plugin-created runtime; hardware plugins own runtime construction and may configure its scheduler or gamepad. Simulation entry points continue to return a SimulationRuntime.

See the deployment tutorial for export, simulation, and hardware workflows, or the Python example for programmatic control.