DEBUNK.IT

AI-powered misinformation detection platform

Analyze claims before they spread.

---

🚀 Live Demo

Try DEBUNK.IT right now — no setup required:

👉 debunkit.onrender.com

⏱️ The free Render instance may take 30–60 seconds to wake up after inactivity. Just wait a moment and it will load.

---

📸 Overview

DEBUNK.IT is a full-stack, AI-assisted fact-checking web platform that analyzes headlines, article URLs, and full text for misinformation. It combines:

• 🤖 Cohere LLM for source-aware AI reasoning
• 🌐 Live web search (RAG) via DuckDuckGo for real-time context
• 📊 TF-IDF + Logistic Regression as a trained local ML fallback
• 🔍 Custom NLP engine for pattern-based offline analysis
• 🔒 Security-first architecture — CSRF, SSRF, XSS, rate limiting

Built for the modern misinformation landscape — sports claims, political statements, health news, viral videos, and general news across Indian and international sources.

---

🎯 Motivation

Misinformation spreads faster than fact-checking can keep up.

DEBUNK.IT was built to close that gap — giving anyone the ability to quickly evaluate news headlines, article URLs, and viral claims using AI-assisted reasoning combined with real-time web evidence. No subscriptions, no black boxes — just transparent verdicts with sources, confidence scores, and clear reasoning.

---

✨ Features

Feature	Description
🔍 Hybrid AI Pipeline	Cohere LLM + ML model + NLP fallback — designed to always provide an evidence-based assessment
🌐 Live RAG Search	Multi-query DuckDuckGo search with fact-check source detection
🔗 URL Article Scraper	Extracts article content and uses it as primary evidence
🛡️ Security-First	CSRF tokens, SSRF blocking, XSS sanitization, rate limiting, hashed passwords
📚 Analysis History	Full searchable/filterable history with verdict statistics per user
📊 Engine Telemetry	Verdict breakdown, confidence averages, mode usage analytics
📋 Report Export	Copy or download full analysis reports as .txt files
🎨 Polished Dark UI	Animated Matrix-style background, dark/light theme, monospace terminal aesthetic
💾 Smart Caching	Results cached per user (6h for URLs, 24h for text)
👤 User Accounts	Register, login, profile management, settings, account deletion

---

📸 Screenshots

Live Scanner — REFUTED verdict at 85% confidence	Login page — animated canvas background
Database Ledger — full analysis history with filters	Engine Telemetry — verdict breakdown & stats

flowchart TD
    A["USER INPUT"] --> B["INPUT VALIDATION LAYER"]
    B -- "URL mode only" --> C["ARTICLE SCRAPER"]
    B -- "Headline or Text mode" --> D["RAG LIVE WEB SEARCH"]
    C --> D
    D --> E["COHERE AI ANALYSIS"]
    E -- "AI unavailable" --> F["LOCAL NLP / ML FALLBACK"]
    E --> G["VERDICT + CONFIDENCE SCORE"]
    F --> G
    G --> H[("DATABASE")]

    classDef yellow fill:#e8c84a,stroke:#333,stroke-width:2px,color:#000;
    classDef red fill:#e05252,stroke:#333,stroke-width:2px,color:#fff;
    classDef green fill:#3ecf7a,stroke:#333,stroke-width:2px,color:#000;
    classDef grey fill:#6f7890,stroke:#333,stroke-width:2px,color:#fff;

    class A,G yellow;
    class B red;
    class C,D,E,F green;
    class H grey;

Loading

Verdict System

Verdict	Meaning
✅ SUPPORTED	Credible evidence confirms the central claim
❌ REFUTED	Credible sources contradict the claim
⚠️ MISLEADING	Partly true but missing context, old media, or misrepresented
🔍 INSUFFICIENT EVIDENCE	Sources don't clearly confirm or deny
🚩 LOW CREDIBILITY	Suspicious language/weak sourcing, can't be directly refuted

---

📈 Project Metrics

Metric	Value
🐍 Python code	~2,000+ lines
🌐 JavaScript	~1,000+ lines
📡 API endpoints	10 endpoints
🔍 Analysis modes	3 (Headline / URL / Text)
🧠 AI verdicts	5 possible outcomes
🔒 Security layers	5 independent layers
👤 Auth system	Full register / login / settings / delete
💾 Caching	Smart freshness rules (6h URL / 24h text)

---

🗂️ Project Structure

Debunkit/
├── debunkit_app.py          # Flask app — all routes & API endpoints
├── config.py                # Config classes (Dev / Prod / Test)
├── requirements.txt         # Python dependencies
├── training.py              # ML model training (TF-IDF + LogReg)
├── dataset.py               # Dataset inspection utility
│
├── Core/
│   ├── ai_engine.py         # Cohere AI + hybrid analysis logic
│   ├── rag_engine.py        # DuckDuckGo live search + article scraper
│   ├── nlp_engine.py        # Local NLP pattern engine (offline fallback)
│   ├── database.py          # SQLAlchemy models + DB helpers
│   └── user_model.py        # User auth model (Flask-Login)
│
├── utils/
│   ├── validators.py        # Input validation + SSRF protection
│   └── sanitizer.py         # XSS / URL injection prevention
│
├── Template/                # Jinja2 HTML templates
│   ├── index.html           # Main dashboard (scanner, history, analytics)
│   ├── login.html
│   ├── register.html
│   ├── profile.html
│   └── settings.html
│
└── Static/
    ├── CSS/                 # style.css, auth.css, profile.css
    ├── Images/              # Logo and assets
    └── JS/
        ├── canvas.js        # Animated falling-words background
        └── main.js          # Full frontend logic (~900 lines)

---

⚙️ Tech Stack

Backend

• Flask — web framework
• Flask-Login — user session management
• Flask-Limiter — rate limiting
• Flask-SQLAlchemy — ORM (SQLite / PostgreSQL)
• Flask-CORS — cross-origin resource sharing

AI / ML

• Cohere API (command-r-08-2024) — primary AI fact-checker
• scikit-learn — TF-IDF + Logistic Regression local model
• DuckDuckGo Search (ddgs) — RAG live web context
• BeautifulSoup4 — article scraping

Frontend

• Jinja2 templates + vanilla JS (~900 lines)
• CSS custom properties — dark/light theme
• HTML5 Canvas — animated background

---

🚀 Quick Start

1. Clone the repository

git clone https://github.com/Code-Breaker-Ctrl/Debunkit.git
cd Debunkit

2. Install dependencies

pip install -r requirements.txt

3. Set up environment variables

Create a .env file in the project root:

COHERE_API_KEY=your_cohere_api_key_here
SECRET_KEY=your_flask_secret_key_here
DATABASE_URL=sqlite:///debunkit.db
FLASK_ENV=development

Get a free Cohere API key at cohere.com

4. (Optional) Train the local ML model

python training.py

Requires news_benchmark_200_clean.csv in the project root.
Skip this step to run with AI-only mode (Cohere).

5. Run the app

python debunkit_app.py

Open http://127.0.0.1:5000 in your browser.

---

🔐 Security Architecture

DEBUNK.IT is built with a layered security approach:

Layer	Protection
validators.py	Input length limits, null byte stripping, URL scheme enforcement, SSRF blocking via DNS resolution
sanitizer.py	XSS prevention — HTML escaping, script tag removal, dangerous protocol blocking (javascript:, data:)
debunkit_app.py	CSRF token validation on every POST / PUT / PATCH / DELETE request
main.js	Sends CSRF token in X-CSRFToken header; blocks non-HTTP source URLs in the DOM
config.py	Rate limiting (10/day guests, 20/min users), password complexity regex, secure session cookies

---

📡 API Endpoints

Method	Endpoint	Description
POST	/analyze	Main analysis endpoint
GET	/api/history	Get user's analysis history
GET	/api/analysis/<id>	Get specific analysis details
GET	/api/search?q=	Search past analyses
GET	/api/stats	Get verdict statistics
POST	/api/clear-database	Delete all user analyses
GET/PUT	/api/user/profile	View or update profile
PUT	/api/user/settings	Update theme, password, preferences
DELETE	/api/user/delete	Delete account and all data
GET	/api/health	Health check

---

🌍 Deployment

Current Deployment

Property	Value
🖥️ Platform	Render
🌐 URL	debunkit.onrender.com
⚡ Status	✅ Production Ready
🗄️ Database	SQLite (upgrade to PostgreSQL planned)
🚀 Server	Gunicorn (multi-worker)

Environment Variables (Production)

FLASK_ENV=production
SECRET_KEY=<strong-random-secret>
COHERE_API_KEY=<your-key>
DATABASE_URL=postgresql://user:pass@host/dbname
SESSION_COOKIE_SECURE=True
RATELIMIT_ENABLED=True

Run with Gunicorn

pip install gunicorn
gunicorn -w 4 -b 0.0.0.0:5000 debunkit_app:app

Docker (coming soon)

# Dockerfile support planned for v3.1

---

⚠️ Limitations

• Not a final truth authority. DEBUNK.IT is an AI-assisted tool, not a replacement for human judgment.
• Breaking news, satire, AI-generated images, and conflicting sources may still require manual verification.
• The local ML model (training.py) is a style classifier — it detects writing patterns, not factual truth.
• AI analysis requires a valid Cohere API key. Without it, the system falls back to local NLP only.
• Search results depend on DuckDuckGo availability — analysis quality may vary for very recent events.

---

🏆 Project Highlights

• ✅ Full-stack AI-powered misinformation detection platform
• ✅ Live public deployment on Render
• ✅ Cohere-powered source-aware fact-checking with domain-specific prompts
• ✅ Hybrid architecture — AI + ML + NLP, always provides an assessment
• ✅ Secure authentication system with CSRF, SSRF, and XSS protection
• ✅ User analytics, scan history, and report export
• ✅ Animated Matrix-style UI with dark/light theme support

---

🗺️ Roadmap

✅ Completed

• Core analysis pipeline (AI + ML + NLP)
• User authentication & accounts
• Analysis history & Engine Telemetry
• Security hardening (CSRF, SSRF, XSS, rate limiting)
• Live deployment on Render

🔜 Planned

• Custom domain (debunkit.tech)
• Docker support
• PostgreSQL migration
• Browser extension
• Image misinformation detection
• Video misinformation detection
• Multilingual support
• Public API with authentication

---

🤝 Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/your-feature)
3. Commit your changes (git commit -m 'Add: your feature')
4. Push to the branch (git push origin feature/your-feature)
5. Open a Pull Request

---

📄 License

This project is licensed under the MIT License.

---

Built by Code-Breaker-Ctrl

DEBUNK.IT v3.0 — AI-powered misinformation detection

⭐ Star this repo if you found it useful!