docs/articles_en/physical-ai/reference/cli.md
The runtime CLI commands use the same schemas as the Python APIs.
The canonical command is physicalai. A shorthand alias pai is also
installed and behaves identically.
physicalai runphysicalai run --config runtime.yaml [--run.duration_s=60]
Arguments:
| Argument | Required | Description |
|---|---|---|
--config | yes | Runtime config YAML |
--run.duration_s | no | Stop after the given duration in seconds |
The same duration limit is available from the current Python API after you construct a runtime directly.
runtime = PolicyRuntime(...)
with runtime:
runtime.run(duration_s=60)
Shell completion scripts can be printed directly from the CLI and sourced in your shell. Completion includes any installed plugin subcommands, such as the studio training commands.
source <(physicalai completion zsh)
source <(pai completion bash)
Training packages can add commands through entry points.
[project.entry-points."physicalai.cli.subcommands"]
fit = "physicalai.cli.fit:register"
validate = "physicalai.cli.validate:register"
test = "physicalai.cli.test:register"
predict = "physicalai.cli.predict:register"
benchmark = "physicalai.cli.benchmark:register"
export = "physicalai.cli.export:register"