Use this flow inside an onboarded project.
python .pf/runtime/bin/pf.py orchestrator-plan create --project-root . --run example-run --title "Example multi-agent run" --apply
python .pf/runtime/bin/pf.py orchestrator-plan validate --project-root . --plan .pf/runs/example-run/orchestrator-plan.yaml
python .pf/runtime/bin/pf.py orchestrator-plan apply --project-root . --plan .pf/runs/example-run/orchestrator-plan.yaml --apply
python .pf/runtime/bin/pf.py orchestrator-plan status --project-root . --run example-run
python .pf/runtime/bin/pf.py worker-launch-prompt create --project-root . --task docs-worker --applyFlat aliases are also available:
orchestrator-plan-createorchestrator-plan-validateorchestrator-plan-applyorchestrator-plan-statusworker-launch-prompt-createruntime-driver-listworker-run-prepareworker-run-startworker-run-collectsupervisor-run
The plan template is templates/orchestrator-task-plan.yaml.
Plans may include:
runtime:
default_driver: manual
supervisor_profile: default
start_policy: manualWorkers may set runtime_driver and depends_on. The default manual driver
does not start a process; use Runtime driver and supervisor quickstart
for the bounded supervised smoke flow.
To reuse the distribution example plan, run from the ProcessForge distribution
root and point --project-root at an onboarded project:
python bin/pf.py orchestrator-plan create --project-root <project-root> --run example-run --title "Example multi-agent run" --answers examples/multi-agent-orchestration/minimal/orchestrator-task-plan.yaml --apply