v0.1 — Initial AI Automation Framework

CleanSlate Protocol v0.1 — Initial Release

First public version of the CleanSlate AI data standardization framework.

Highlights

• CleanSlate Protocol (CSP) — A three-layer open standard for AI data preprocessing

  • Layer 1: Raw Extraction (source preservation + SHA-256 checksum)
  • Layer 2: Semantic Structure (heading detection, table analysis, entity extraction)
  • Layer 3: Verification (Merkle hash tree, Ed25519 signatures, timestamps)

• PDF Engine v0.2 — Production-grade PDF parsing engine

  • Font-clustering heading detection with hierarchy inference
  • Spatial column analysis for table extraction
  • CJK-Latin mixed text spacing repair
  • OCR artifact cleanup
  • Overall accuracy: 82/100 (up from 15/100 in v0.1)

• 10+ Format Support — PDF, DOCX, XLSX, PPTX, HTML, CSV, TXT, JSON, PNG, JPEG, Markdown

• CSP Playground — Browser-based interactive converter

  • Drag-and-drop file upload
  • Three output modes: CSP JSON / Markdown / Raw JSON
  • Real-time quality metrics display
  • 100% browser-local processing — zero uploads, zero servers

• Professional Landing Page — Protocol showcase with architecture diagrams, code examples, ecosystem overview, and competitive comparison

Technical Stack

• React 19 + TypeScript 5.6
• Tailwind CSS 4 + shadcn/ui
• Vite 7 build system
• pdfjs-dist for PDF processing
• Framer Motion for animations

Getting Started

git clone https://github.com/andi586/cleanslate-pdf-engine.git
cd cleanslate-pdf-engine
pnpm install
pnpm dev

Open http://localhost:3000 in your browser.

Documentation

• README.md — Project overview
• ARCHITECTURE.md — Technical deep-dive
• ROADMAP.md — Development roadmap
• CONTRIBUTING.md — Contribution guidelines

License

MIT License — free for personal and commercial use.

---

CleanSlate Protocol — The standardization layer every AI agent needs.