How siriacid/right-to-intervene works
Right to Intervene is a synthetic Python research benchmark for comparing production-control policies when equipment health is uncertain and human reviews are delayed. Researchers run paired trials across policies and disruption scenarios; the simulator models job flow, health observations, permission decisions, reviews, and a hard quality interlock, then the runner summarizes outcomes and paired effects. A separate sensitivity runner varies review budgets and delays. The older v0.1 prototype is retained as a historical benchmark. The sampled code shows the current runners’ calls and output generation; detailed simulator internals are only partially excerpted.
Main entry points: authority_v2/experiment.py Paired Experiment Runner · sensitivity.py Review Sensitivity Runner · factory_v0_1/experiment.py Factory v0.1 Benchmark · simulator.py Trial Simulator
Components 8
Current Simulation
- Trial Simulator
experiments/authority_v2/simulator.py
Experiment Evaluation
- Paired Experiment Runner
experiments/authority_v2/experiment.py - Review Sensitivity Runner
experiments/authority_v2/sensitivity.py - Trial and Summary Outputs
experiments/authority_v2/results/summary.csv - Sensitivity Results
experiments/authority_v2/results/review_sensitivity.csv
Historical Prototype
- Factory v0.1 Benchmark
experiments/factory_v0_1/experiment.py - Legacy Summary
experiments/factory_v0_1/results/summary.csv
Other components
- Researcher
Connections 8
- Researcher → Paired Experiment Runner: runs trials
README.md - Paired Experiment Runner → Trial Simulator: runs each trial
experiments/authority_v2/experiment.py - Paired Experiment Runner → Trial and Summary Outputs: writes results
experiments/authority_v2/experiment.py - Researcher → Review Sensitivity Runner: varies review settings
README.md - Review Sensitivity Runner → Trial Simulator: runs trials
experiments/authority_v2/sensitivity.py - Review Sensitivity Runner → Sensitivity Results: writes results
experiments/authority_v2/sensitivity.py - Researcher → Factory v0.1 Benchmark: runs historical benchmark
README.md - Factory v0.1 Benchmark → Legacy Summary: writes summary
experiments/factory_v0_1/experiment.py