CLI Reference#

Preview: The CLI is a planned API. The commands below document the target design.

The runtime CLI commands use the same schemas as the Python APIs.

physicalai run#

physicalai run --config runtime.yaml [--duration-s 60]

Arguments:

Argument

Required

Description

--config

yes

Runtime config YAML

--duration-s

no

Stop after the given duration in seconds

The equivalent Python call is shown below.

PolicyRuntime.from_config("runtime.yaml").run(duration_s=60)

physicalai infer#

(Planned API — interface may change.)

physicalai infer --config inference.yaml

Arguments:

Argument

Required

Description

--config

yes

Inference config YAML

physicalai serve#

physicalai serve --config server.yaml --host 0.0.0.0 --port 8080

Use this command when inference should run remotely instead of on the robot host.

Plugin Commands#

Training packages can add commands through entry points.

[project.entry-points."physicalai.cli.subcommands"]
fit = "physicalai.train.cli:register_fit"
validate = "physicalai.train.cli:register_validate"
test = "physicalai.train.cli:register_test"
predict = "physicalai.train.cli:register_predict"
export = "physicalai.train.cli:register_export"