@6overn/rdl — AI Governance Engine

Govern AI output with verified facts. Every claim extracted, verified against reality, and scored.

Patent: US10042928 — Rule Description Logic (deontic-epistemic coupling)

Install

npm install @6overn/rdl

Quick Start

import { govern } from "@6overn/rdl";

const result = await govern("The market is worth $50 billion.", {
  mode: "flag"
});

console.log(result.score);           // 0.0 - 1.0 accuracy score
console.log(result.claims);          // extracted claims with verdicts
console.log(result.authenticity);    // badge: authentic/mixed/performing/governed

What It Does

1. Extracts typed claims from any text (dollars, dates, URLs, entities, citations, statistics)
2. Verifies each claim against a 5-tier grounding cascade (32 sources)
3. Scores the output with 7 trust badges
4. Enforces governance verdicts (12 modes: block, correct, flag, audit, explain, ...)

Badges

• ◆ Authentic — honest, sourced, hedges appropriately
• ◇ Mixed — some claims verified, some bare
• ✕ Performing — high specificity, zero sources (hallucination risk)
• ✦ Governed — corrections applied, output improved

12 Enforcement Modes

block | correct | flag | audit | coach | redact | annotate | source | strict | compare | score_only | explain

Documentation

See RDL for AI Governance

License

Proprietary. See LICENSE.md. Patent US10042928.