CLI reference
Generated from the command parser. Start with pixi run setup.
Human diagnostics use stderr; --json produces structured stdout. --plain disables styling.
Shell completion: inside pixi shell, run eval "$(dwarl completion bash)" (also supports zsh and fish).
dwarl
usage: dwarl [-h] [--plain] [--json] [--verbose] COMMAND ...
Train, test and deploy a local planner for your robot.
options:
-h, --help show this help message and exit
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Commands:
COMMAND
setup Get started: try an example, integrate a robot or
continue
train Train a policy
test Evaluate an exact checkpoint
deploy Install a compatible policy bundle
status Inspect progress
next Show the next useful step
runs List, inspect or watch training runs
report Summarize or open a navigation report
doctor Check the environment or a running simulator integration
robot Create or inspect a robot integration
stop Interrupt an owned training process
completion Print shell-completion setup
model-benchmark Replay recorded ONNX inputs
Getting started: setup · robot init
Daily workflow: train · test · deploy
Inspection: status · next · runs · report · doctor
Start: pixi run setup
Next step: pixi run dwarl next
Docs: https://nobleo.github.io/dwarl/
dwarl setup
usage: dwarl setup [-h] [--plain] [--json] [--verbose] [--robot-ws ROBOT_WS]
[--packages PACKAGES [PACKAGES ...]] [--example] [--yes]
[--no-build]
Get started: try an example, integrate a robot or continue
options:
-h, --help show this help message and exit
--robot-ws ROBOT_WS robot repository containing src/
--packages PACKAGES [PACKAGES ...]
integration packages to build
--example select the included diffbot
--yes use complete, unambiguous saved/detected inputs
--no-build save selection without building ROS
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run setup
dwarl train
usage: dwarl train [-h] [--plain] [--json] [--verbose] [--config CONFIG]
[--set PATH=VALUE] [--print-config] [--dry-run]
[--preset {smoke,navigation}] [--num-envs NUM_ENVS]
[--seed SEED] [--max-iterations MAX_ITERATIONS]
[--device DEVICE] [--name NAME] [--resume RESUME]
[--detached] [--show]
[robot]
Train a policy
positional arguments:
robot integration/profile (default: setup selection)
options:
-h, --help show this help message and exit
--config CONFIG YAML configuration (replaces ./dwarl.yaml)
--set PATH=VALUE advanced schema override
--print-config print resolved configuration and exit
--dry-run inspect selected inputs without executing
--preset {smoke,navigation}
explicit starter recipe
--num-envs NUM_ENVS
--seed SEED
--max-iterations MAX_ITERATIONS
--device DEVICE
--name NAME
--resume RESUME
--detached
--show
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run train --help
dwarl test
usage: dwarl test [-h] [--plain] [--json] [--verbose] [--config CONFIG]
[--set PATH=VALUE] [--print-config] [--dry-run] [--run RUN]
[--checkpoint CHECKPOINT] [--sim {mujoco,gazebo,both}]
[--suite {benchmark,validation,test}] [--case CASE]
[--repeats REPEATS] [--controllers CONTROLLERS]
[--profile-compute] [--show]
Evaluate an exact checkpoint
options:
-h, --help show this help message and exit
--config CONFIG YAML configuration (replaces ./dwarl.yaml)
--set PATH=VALUE advanced schema override
--print-config print resolved configuration and exit
--dry-run inspect selected inputs without executing
--run RUN run directory (default: selected robot's navigation
run, then smoke)
--checkpoint CHECKPOINT
latest, best or iteration (default: latest)
--sim {mujoco,gazebo,both}
--suite {benchmark,validation,test}
--case CASE
--repeats REPEATS
--controllers CONTROLLERS
dwarl,mppi,dwb (Gazebo)
--profile-compute
--show
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run test --help
dwarl deploy
usage: dwarl deploy [-h] [--plain] [--json] [--verbose] [--config CONFIG]
[--set PATH=VALUE] [--print-config] [--dry-run]
[--run RUN] [--checkpoint CHECKPOINT]
[robot]
Install a compatible policy bundle
positional arguments:
robot integration/profile (default: setup selection)
options:
-h, --help show this help message and exit
--config CONFIG YAML configuration (replaces ./dwarl.yaml)
--set PATH=VALUE advanced schema override
--print-config print resolved configuration and exit
--dry-run inspect selected inputs without executing
--run RUN run directory (default: selected robot's navigation
run, then smoke)
--checkpoint CHECKPOINT
latest, best or iteration (default: latest)
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run deploy --help
dwarl status
usage: dwarl status [-h] [--plain] [--json] [--verbose] [--robot ROBOT]
[--run RUN]
Inspect progress
options:
-h, --help show this help message and exit
--robot ROBOT inspect an integration without changing the saved selection
--run RUN inspect a particular run
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run dwarl status
dwarl next
usage: dwarl next [-h] [--plain] [--json] [--verbose] [--robot ROBOT]
[--run RUN]
Show the next useful step
options:
-h, --help show this help message and exit
--robot ROBOT inspect an integration without changing the saved selection
--run RUN inspect a particular run
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run dwarl next
dwarl runs
usage: dwarl runs [-h] [--plain] [--json] [--verbose] [--interval INTERVAL]
[{list,show,watch}] [run]
List, inspect or watch training runs
positional arguments:
{list,show,watch}
run run directory (default: selected run)
options:
-h, --help show this help message and exit
--interval INTERVAL watch refresh interval in seconds
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run dwarl runs watch
dwarl report
usage: dwarl report [-h] [--plain] [--json] [--verbose] [--run RUN] [--open]
[directory]
Summarize or open a navigation report
positional arguments:
directory test directory (default: latest test of selected run)
options:
-h, --help show this help message and exit
--run RUN
--open open the local HTML report in a browser
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run dwarl report --open
dwarl doctor
usage: dwarl doctor [-h] [--plain] [--json] [--verbose]
[--sim {mujoco,gazebo}] [--robot ROBOT] [--live]
[--timeout TIMEOUT]
Check the environment or a running simulator integration
options:
-h, --help show this help message and exit
--sim {mujoco,gazebo}
--robot ROBOT
--live launch a managed simulator and verify its interfaces
--timeout TIMEOUT
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run dwarl doctor --sim gazebo --live
dwarl robot
usage: dwarl robot [-h] [--plain] [--json] [--verbose] ACTION ...
Create or inspect a robot integration
positional arguments:
ACTION
check Check the selected robot
view View the selected robot
init Create or update a robot-owned integration
options:
-h, --help show this help message and exit
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run dwarl robot init
dwarl robot check
usage: dwarl robot check [-h] [--plain] [--json] [--verbose] [robot]
positional arguments:
robot integration/profile (default: setup selection)
options:
-h, --help show this help message and exit
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
dwarl robot view
usage: dwarl robot view [-h] [--plain] [--json] [--verbose] [robot]
positional arguments:
robot integration/profile (default: setup selection)
options:
-h, --help show this help message and exit
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
dwarl robot init
usage: dwarl robot init [-h] [--plain] [--json] [--verbose]
[--answers ANSWERS] [--name NAME] [--set PATH=VALUE]
[--yes] [--update] [--profile-only]
[--license LICENSE] [--maintainer MAINTAINER]
[--email EMAIL] [--gazebo] [--sim-package SIM_PACKAGE]
[--sim-launch SIM_LAUNCH] [--arg-map ARG=SIM_ARG]
[directory]
positional arguments:
directory
options:
-h, --help show this help message and exit
--answers ANSWERS input robot profile YAML; relative assets resolve
beside this file
--name NAME ROS package name (default: directory name)
--set PATH=VALUE
--yes non-interactive; required inputs must be supplied
--update apply displayed changes to generated files
--profile-only create a profile without ROS package scaffolding
--license LICENSE license for the generated robot package
--maintainer MAINTAINER
--email EMAIL
--gazebo wrap an existing robot-only simulator launch
--sim-package SIM_PACKAGE
--sim-launch SIM_LAUNCH
--arg-map ARG=SIM_ARG
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
dwarl stop
usage: dwarl stop [-h] [--plain] [--json] [--verbose] run
Interrupt an owned training process
positional arguments:
run
options:
-h, --help show this help message and exit
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run dwarl stop runs/<run>
dwarl completion
usage: dwarl completion [-h] [--plain] [--json] [--verbose] [{bash,zsh,fish}]
Print shell-completion setup
positional arguments:
{bash,zsh,fish}
options:
-h, --help show this help message and exit
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run dwarl completion bash
dwarl model-benchmark
usage: dwarl model-benchmark [-h] [--plain] [--json] [--verbose]
[--output OUTPUT] [--threads THREADS]
[--iterations ITERATIONS] [--warmup WARMUP]
bundle observations
Replay recorded ONNX inputs
positional arguments:
bundle
observations
options:
-h, --help show this help message and exit
--output OUTPUT
--threads THREADS
--iterations ITERATIONS
--warmup WARMUP
Output:
--plain plain, non-animated output
--json structured result on stdout
--verbose show detailed diagnostics and traceback
Example: pixi run dwarl model-benchmark --help
Default configuration
Robot profile schema
See the robot profile reference.