🔬 DetectAI — Full-Stack AI Text Detector

Detect whether text was written by a human or an AI. Supports plain text, URLs, and file uploads. Built with React + Vite (frontend), Express + SQLite (backend), powered by Claude AI.

---

📁 Project Structure

ai-detector/
├── backend/
│   ├── server.js          # Express entry point
│   ├── db.js              # SQLite database (better-sqlite3)
│   ├── routes/
│   │   ├── detect.js      # POST /api/detect/text|url|file
│   │   └── history.js     # GET/DELETE /api/history
│   ├── .env.example       # Copy to .env and fill in
│   ├── Dockerfile
│   └── package.json
│
├── frontend/
│   ├── src/
│   │   ├── App.jsx
│   │   ├── main.jsx
│   │   ├── index.css
│   │   ├── utils/api.js          # Axios API helpers
│   │   ├── components/
│   │   │   ├── Navbar.jsx
│   │   │   ├── Results.jsx       # Detection results display
│   │   │   └── UI.jsx            # Shared components
│   │   └── pages/
│   │       ├── Detect.jsx        # Main detection page
│   │       ├── History.jsx       # Scan history
│   │       ├── ScanDetail.jsx    # Single scan detail
│   │       └── Stats.jsx         # Statistics dashboard
│   ├── vite.config.js
│   ├── Dockerfile
│   ├── nginx.conf
│   └── package.json
│
├── docker-compose.yml
└── README.md

---

🚀 Quick Start (Local Development)

1. Clone and install dependencies

# Backend
cd backend
npm install

# Frontend
cd ../frontend
npm install

2. Configure backend environment

cd backend
cp .env.example .env

Edit .env:

PORT=4000
ANTHROPIC_API_KEY=sk-ant-your-key-here   # Get from console.anthropic.com
FRONTEND_URL=http://localhost:5173

3. Start backend

cd backend
npm run dev      # uses nodemon for auto-reload
# or
npm start

Backend runs at: http://localhost:4000

4. Start frontend

cd frontend
npm run dev

Frontend runs at: http://localhost:5173

---

🐳 Docker (Production)

# At project root
cp backend/.env.example backend/.env
# Edit backend/.env with your ANTHROPIC_API_KEY

docker-compose up --build

• Frontend: http://localhost:5173
• Backend API: http://localhost:4000

---

🔌 API Reference

Detect Text

POST /api/detect/text
Content-Type: application/json

{ "text": "Your text here..." }

Detect from URL

POST /api/detect/url
Content-Type: application/json

{ "url": "https://example.com/article" }

Detect from File

POST /api/detect/file
Content-Type: multipart/form-data

file: <file upload>   (.txt, .md, .html)

Get History

GET /api/history?page=1&limit=20

Get Single Scan

GET /api/history/:id

Delete Scan

DELETE /api/history/:id

Stats

GET /api/history/stats

Health Check

GET /api/health

---

🗄️ Database Schema

SQLite database is stored at backend/data/detector.db.

scans table — stores every detection result:

Column	Type	Description
id	TEXT (UUID)	Primary key
created_at	DATETIME	Timestamp
source_type	TEXT	text, url, or file
source_ref	TEXT	URL or filename
input_text	TEXT	First 500 chars of input
word_count	INTEGER	Total words
verdict	TEXT	AI, Human, or Mixed
ai_prob	REAL	0–100 probability
confidence	TEXT	Low/Medium/High/Very High
summary	TEXT	Analysis summary
signals	TEXT	JSON: perplexity/burstiness/etc
ai_flags	TEXT	JSON array
human_flags	TEXT	JSON array
phrases	TEXT	JSON array of flagged phrases
ip_address	TEXT	Client IP

stats table — running totals for the dashboard.

---

⚙️ Tech Stack

Layer	Tech
Frontend	React 18, Vite, React Router
Backend	Node.js, Express 4
Database	SQLite via better-sqlite3
AI	Anthropic Claude (claude-sonnet-4)
Web scraping	Axios + Cheerio
File uploads	Multer
Security	Helmet, express-rate-limit, CORS
Deployment	Docker + Nginx

---

🔒 Security Notes

• Rate limiting: 30 requests/minute per IP on /api/detect
• File uploads capped at 5MB
• Input text truncated to 8000 characters before sending to Claude
• CORS restricted to FRONTEND_URL
• Helmet sets secure HTTP headers

---

📝 License

MIT