Deontica Sandbox

One canonical rule. Five coordinated artifacts.

Deontica structures transportation regulations into machine-readable rules and emits OpenSCENARIO scenarios, Python validators, smooth ML loss functions, JSON constraints, and violation functions — all from one canonical source. Below are five sample rules from real regulations (TX Transportation Code Ch. 545), rendered as they'd ship in your bundle.

Download a sample bundleTry it on your regulations →

Try it live — R-079 speed limit

The actual ML-mode violation primitive from dt_eval_categories.violation_speed, running in your browser via Pyodide. Slide the predicted speed past the threshold and watch the smooth severity curve respond — this is the gradient signal an AV planner trains against.

18.0m/s
13.4m/s
Severity vs predicted speed
ML (smooth)V&V (binary)
Click Run to load Pyodide and render the live severity curve.
Math runs in your browser via Pyodide (0.27.5). The snippet matches dt_eval_categories.violation_speed from the bundled Python runtime — same primitives the downloadable bundle uses.

Showcase rules

Each card expands to show the same rule rendered five different ways. Tab between formats to see how the same canonical rule becomes a scenario, a validator, a loss function, a constraint, and a violation function.

Sample rules drawn from publicly-available statute text. Production decompositions are typed by curated atom semantics and validated against the same export-engine validator the downloadable bundle uses.