robo CLI

robo is the agent's only way to act. It is one standard-library Python file that sends one JSON request per call to the episode server at $ROBOUSE_SOCKET. The runner puts it first on the agent's PATH; pip install robouse also installs it.

$ robo help
`robo`: the only way an agent touches the robot. Standard library only.

Finds the episode socket from $ROBOUSE_SOCKET (set by the runner) and sends one JSON request.

  robo info                          action space, skills, step budget
  robo observe [--image] [--camera C] robot and object state; --image also saves a camera PNG and prints its path
  robo act A1 A2 ... [--repeat N]    apply one low-level action N times (N <= 50)
  robo move-to X Y Z [--grip G]      move the hand toward a point (skill; may be disabled per task)
  robo grip G [--steps N]            hold position and set the gripper (+1 close, -1 open)
  robo skill NAME [ARG ...]          run a named high-level skill listed by `robo info` (only some tasks have them)
  robo done ["message"]              end the episode and ask for scoring
  robo give-up ["message"]           end the episode without claiming success
Add --json to any command for raw JSON output.

Commands#

Command	Notes
`robo info`	Action names and bounds, available skills, steps used and budget. On vision tasks also the visible fields and each camera's size and projection matrix.
`robo observe`	State as numbers. `--image` also saves a PNG from the task's camera into the workspace's `observations/` and prints its path. On vision tasks it saves one image per camera on every call; `--camera C` picks one of them.
`robo act A1 A2 ...`	One low-level action, applied `--repeat N` times. N is clipped to 50, or to the task's `max_repeat`.
`robo move-to X Y Z`	Skill: moves the hand toward a point. `--grip G` sets the gripper during the move; stops within `--tol` metres (default 0.01) or after `--max-steps` steps (default 100, at most 150).
`robo grip G`	Skill: holds position and sets the gripper for `--steps N` steps (default 15).
`robo skill NAME`	A named skill, when `robo info` lists one. No task in the 0.1.0 package has them.
`robo done "message"`	Ends the episode and asks for scoring.
`robo give-up "reason"`	Ends the episode without claiming success.

Skills are sequences of low-level actions run by the server, and each simulated step counts against the budget. A task can turn skills off (skills: false), leaving only robo act. Without an episode, commands print an error and exit 1: error: no episode: ROBOUSE_SOCKET is not set, or error: episode server is not running (episode may have finished).

Protocol#

One JSON object per connection, newline-terminated, for example {"op": "act", "action": [0, 0, -1, 1], "repeat": 5}. The response is {"ok": true, "result": {...}} or {"ok": false, "error": "..."}. Operations: info, status, observe, act, move_to, grip, skill, done, give_up and shutdown. robo does not expose shutdown, which ends the episode as agent_exited. There is no reset, teleport or state-setting operation.