VoxCivicAI

AI-Powered Policy Comment Review Platform

VoxCivicAI makes government regulations accessible to everyone. It transforms dense policy documents into plain language, enables inline public commenting, and provides agencies with powerful analytics to process feedback at scale.

---

Features

🔍 Plain-Language Translation

Automatically converts complex regulatory language into clear, accessible text using fine-tuned LLMs. Side-by-side comparison with readability scoring ensures quality.

💬 In-Document Commenting

Attach comments directly to specific passages in policy documents. See existing feedback in context and engage in targeted discussions.

🛡️ Mass-Campaign & Fraud Detection

Identifies duplicate submissions, computer-generated comments, and suspicious patterns. Integrates reCAPTCHA and supports misattribution reporting.

📊 Comment Summarization & Clustering

AI-generated summaries and topic clustering using semantic embeddings and hierarchical topic modeling (hLDA). Quickly surface major themes from thousands of comments.

📈 Analytics Dashboard

Real-time insights: total comments, duplicates, sentiment analysis, topic distribution, time-series trends, and top keywords. Slice by commenter type or topic.

🔥 Document Heat-Map Analytics

Visualize reader engagement and comment density. Identify confusing or contentious sections at a glance.

✅ Responsible AI

Human-in-the-loop review, model confidence transparency, and analyst feedback loops for corrections.

---

Tech Stack

Backend

• Python 3.11+ with FastAPI
• LangGraph for AI agent orchestration
• OpenAI GPT-4o for translation and summarization
• ChromaDB for vector similarity search
• SQLite (dev) / PostgreSQL (prod)

Frontend

• React 18 with TypeScript
• Vite build tooling
• Tailwind CSS + shadcn/ui components
• TanStack Query for data fetching
• Recharts for visualizations

---

Getting Started

Prerequisites

• Node.js 18+
• Python 3.11+
• Docker (optional)

Frontend Setup

cd frontend
npm install
npm run dev

Backend Setup

cd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
uvicorn main:app --reload

Environment Variables

Copy .env.example to .env and configure:

OPENAI_API_KEY=your_key_here
DATABASE_URL=sqlite:///./voxcivicai.db

---

Development

Local Ollama (Optional)

A local Ollama instance can be used for development:

• Base URL: http://localhost:11434
• Models: qwen2.5:7b (chat), nomic-embed-text (embeddings)

Code Standards

• Python: PEP 8, Black formatter, Ruff linting, Google-style docstrings
• TypeScript: ESLint + Prettier, strict mode, functional components

Testing

# Backend
cd backend && pytest

# Frontend
cd frontend && npm test

---

Project Structure

voxcivicai/
├── frontend/          # React TypeScript app
│   ├── src/
│   │   ├── components/
│   │   ├── pages/
│   │   ├── hooks/
│   │   └── api/
│   └── package.json
├── backend/           # FastAPI Python app
│   ├── agents/        # LangGraph AI agents
│   ├── api/           # API routes
│   ├── models/        # Pydantic models
│   └── services/      # Business logic
└── docs/              # Documentation

---

License

MIT

---

Contributing

Contributions welcome! Please read our contributing guidelines and submit PRs against the main branch.