CLI#
The CLI is a thin wrapper over the same config APIs used by Python.
The canonical command is physicalai. A shorthand alias pai is also
installed and behaves identically.
physicalai run --config runtime.yaml --run.duration_s=60
Equivalent Python control flow:
runtime = PolicyRuntime(...)
with runtime:
runtime.run(duration_s=60)
Shell completion is available through jsonargparse + shtab. Print a shell
completion script with the completion command and source or install it in
your shell. When training packages are installed, their plugin subcommands are
included in completion automatically.
source <(physicalai completion zsh)
source <(pai completion zsh)
Runtime Commands#
Command |
Purpose |
|---|---|
|
Runs a policy on robot hardware. |
Training Commands#
Training commands should come from training packages or entry-point plugins.
[project.entry-points."physicalai.cli.subcommands"]
fit = "physicalai.cli.fit:register"
benchmark = "physicalai.cli.benchmark:register"
export = "physicalai.cli.export:register"
Importing physicalai should not pull in training dependencies.