rules-as-code

French tax and benefit system for OpenFisca

OpenFisca core engine. See other repositories for countries-specific code & data.

A user-friendly web-based tool for Rules as Code.

define, compose, and evaluate your business rules in Elixir

A pure, declarative language for business rules.

Start modelling the tax and benefit system of your country in a few minutes.

This project is being continued by Legalese:

Deterministic guardrails for AI agents — the LLM proposes, your rules dispose. A sub-microsecond, JIT-compiled rule engine in Rust, with a visual Studio.

Explore legislation formulas and parameters.

bamboo firewall, firewall, iptables, firewall distributed, host base firewall

Nomos — a programming language for legal reasoning. Typed rules with jurisdiction and validity dates, LLM-powered fact extraction, defeasible logic, proof trees that cite statutes and cases. Apache-2.0. Experimental.

An implementation of Defeasible Deontic Logic in Answer Set Programming

REST server + UI for Rulette

bamboo firewall, firewall, iptables, firewall distributed, host base firewall

Modèle Publicodes pour le calcul des aides vélo en France

L4 - Rules as Code - Legal language syntax highlighting for Highlight.js

GovSCH is an Open-Source Schema for Authoring Cybersecurity & AI Governance Documents

188 DIFC/ADGM/SICC judgments coded under a two-grader-type protocol (39 LLM + 149 regex; provenance per entry). 12 Catala rule modules, 7 case traces, pre-registered with H1–H8 stop rules. External correlate ρ=+0.32 vs appeal status. Reproducible via Docker.

Tax law as a compile target: an LLM compiles the Income Tax Act into a deterministic OpenFisca package — provenance-pinned to verbatim statute, gate-verified to the cent against CRA's published payroll formulas, ambiguity refused. No model at runtime. Canada federal withholding, 2026.

Levee autoformalizes your rulebook (business logic, domain constraints, policy docs) into Lean 4 specs, then enforces them against AI agents at runtime. Formal verification for agentic AI: rules an LLM agent can’t talk its way around.