How I Built a 100% AI-Generated Document Translator with WCAG AAA & Zero-Data Retention

A developer built a full-stack document translator entirely through AI-driven 'Vibe Coding,' converting legal and bureaucratic texts into everyday Portuguese while enforcing WCAG AAA accessibility and LGPD compliance. The project, called Real-World Document Translator, uses a FastAPI backend with PII sanitization and Google Gemini API integration, and can run in a mock mode for offline testing.

Overview:A full-stack, production-grade web application created entirely via AI/Vibe Coding. It converts complex legal terms, fine print, and bureaucratic bills into everyday Portuguese while enforcing WCAG AAA accessibility and strict LGPD compliance. Recently, I set out to test the limits of Vibe Coding by building an end-to-end software product using AI-driven development. The result is the Real-World Document Translator Tradutor de Documentos da Vida Real . Its goal is to bridge the accessibility gap for elderly individuals, people with low vision, and folks with limited literacy—allowing anyone to understand legal and financial contracts without needing a lawyer. Translating legal documents through AI presents two major technical hurdles: | Challenge | Impact / Risk | |---|---| Privacy & LGPD | Raw documents contain sensitive PII CPFs, bank accounts, emails . Standard regex fails due to OCR noise broken numbers, homoglyphs . | Accessibility | Most web tools ignore low-vision users and the elderly, lacking high contrast, scalable typography, and native screen reader tools. | Through iterative prompting and AI engineering, we designed a zero-trust, highly accessible architecture. .env variables. User ───▶️ Accessible Frontend HTML/CSS/JS │ ▼ Secure BFF FastAPI ├── File Validation Magic Bytes ├── PII Sanitization OCR-Resilient └── API Key Guard .env │ ▼ Google Gemini API / Mock AI Building this project 100% through AI was a major technical milestone. It proved that Vibe Coding isn't just about speeding up syntax—it's about directing AI to construct secure, accessible, and architecturally sound systems . Key takeaways from this build: Want to test it locally? Follow these simple steps using Python 3.10+: bash 1. Clone repo & install dependencies pip install -r backend/requirements.txt 2. Run the application python backend/server.py 3. Open http://127.0.0.1:8000 http://127.0.0.1:8000 in your browser Note: If no Gemini API key is configured, the server automatically defaults to Smart Mock AI mode for instant offline testing