CLI
boltrt [--task <name>] --robot <name|yaml> --policy <yaml> --backend <mujoco|real|yaml>
| Option | Meaning |
|---|---|
--task tracking | Select a task; default is the base loop |
--motion <npz|tcp://host:port> | Tracking reference (tracking task only) |
--validate | Check configs, assets, observations and ONNX bindings; start nothing |
--start | Enter the policy and play the motion at startup |
--rate <hz> | Policy rate (default 50) |
--steps <n> / --run-once | Stop after n ticks / after one pass of the motion |
--headless | No window, no real-time pacing |
--no-keyboard | Do not take over the terminal |
--record <jsonl> | Per-tick state, observations, model I/O, actions and commands |
--timings <csv> | Per-tick policy and inference latency |
--events <json> | Scripted key events for unattended runs |
--ort-threads <n> | ONNX Runtime threads (default 1) |
--init-steps <n> | Ticks used by the init mode to reach the default pose (default 500) |
--robot-model <xml|hf://...> | Override the robot model |
--config-root <dir> | Directory of builtin robot and backend configs |
--print-config | Print the resolved configuration |
--list-robots, --list-backends, --list-observations, --list-tasks | List registered names |