m/market labResearch, with a record.
Source on GitHub ↗

Human and Rule decisions

  • runner.py: PolicyDecisionRunner for caller-driven Controlled execution.
  • live/runner.py: LivePolicyDecisionRunner for background Human/Rule execution.
  • live/owner.py: short policy commands and observation capture on the App owner.
  • live/program.py: TrustedPolicyProgram, run in the Rule child; Human has no automatic policy child.
  • policies.py: built-in policies and the external JSON-policy adapter.
  • model_file.py: the model method: bars, the target-order rule and the model process.
  • model_worker.py: runs in the model's own interpreter; loads the user's file (stdlib only).
  • models.py, contracts.py: observations, configurations and policy contracts.
  • controls.py, handoffs.py, scheduling.py: manual control and policy authority changes.
  • invocation.py: calling a policy without owning its financial transaction.

The Live runner coordinates shared experiment infrastructure. It is not the Agent runner and does not contain the user Agent's model/tool loop. See decisions/agent for that method's App-side integration.

Maintained source: sim_platform/decisions/policy/README.md