Standard-Agnostic Compliance Checker

Feed it any safety standard. Feed it any work product. Get a compliance report.

An 8-layer algorithm that checks any document against any standard —
no hardcoded rules, no domain lock-in, no manual cross-referencing.

What it does Give it a safety standard (JSON or Markdown) and a work product (DOCX, PDF, TXT, JSON, MD). It runs eight analysis layers and tells you exactly where you're compliant, where you're not, and what's missing.

The engine is standard-agnostic. It auto-discovers the schema of whatever you throw at it — ISO 26262, IEC 61508, DO-178C, ASPICE, or a standard that doesn't exist yet.

What it looks like

The 8-layer pipeline

#LayerMethod1Node CoverageDecision tree on requirement node existence2Content AlignmentTF-IDF + cosine similarity3Semantic DepthRatcliff/Obershelp fuzzy matching4Concept CoverageSet intersection on discovered concept tags5Reference IntegrityGraph BFS on cross-reference links6Method/Practice AuditRisk-level-aware method matching7Traceability ChainDirected graph walk on dependency paths8Gap Analysis + RiskEnsemble classifier with risk-weighted scoring
Quick start bash# Clone git clone https://github.com/jherrodthomas/compliance-checker-algo.git cd compliance-checker-algo

Install Python dependencies (standard library only — no pip install needed)

Run with synthetic example data

python agnostic_engine.py examples/synthetic_standard examples/synthetic_artifact.json
--meta examples/synthetic_meta.json

Run with your own standard and work product

python agnostic_engine.py /path/to/your/standard /path/to/your/artifact.docx --format all
Output formats --format json → Structured JSON report --format docx → Word document with styled sections --format pdf → PDF with visual compliance dashboard --format html → Interactive HTML report --format all → All of the above
Using your own standards

Standards are copyrighted. This repo includes only synthetic example data. See STANDARDS_NOTICE.md for where to purchase real standards.

The engine accepts standards as JSON files following this schema: json{ "part": "3", "title": "Concept phase", "clauses": [ { "section": "6.1", "title": "Hazard analysis", "type": "requirement", "text": "A hazard analysis shall be performed...", "tags": ["HARA", "hazard analysis"], "asil_level": "all", "notes": ["..."], "cross_references": ["5.1"] } ] } Or as Markdown files (converted from PDF via --convert flag). An optional compliance_meta.json provides schema hints to accelerate discovery — but the engine works without it.
Project structure ├── agnostic_engine.py # Core 8-layer compliance engine ├── artifact_ingester.py # Normalizes DOCX/PDF/MD/TXT/JSON artifacts ├── markdown_parser.py # Converts Markdown standards to engine JSON ├── report_generator.py # JSON report output ├── report_generator_docx.py # Word document report output ├── report_generator_html.py # Interactive HTML report output ├── ISO26262_Checker.py # Legacy ISO 26262-specific checker ├── compliance_meta.json # Example meta-config (schema hints) ├── examples/ # Synthetic test data (no proprietary content) │ ├── synthetic_standard/ # Fictional 4-part safety standard │ ├── synthetic_artifact.json # Fictional work product │ └── synthetic_meta.json # Meta-config for the synthetic standard ├── main.js # Electron desktop app entry ├── renderer/ # Desktop app UI └── STANDARDS_NOTICE.md # Licensing info for real standards
Desktop app The project includes an Electron wrapper (main.js, renderer/, preload.js) that packages the engine as a desktop application. To run it: bashnpm install npm start Build scripts for Windows are in build_app.bat / build_engine.bat.
License MIT — see LICENSE. The compliance engine code is open source. The standards it checks against are not. See STANDARDS_NOTICE.md.

Built by Jherrod Thomas