Luffy AI - Hybrid Fake News Detector 🕵️‍♂️💻

Luffy AI is a high-performance, hybrid fake news detection platform consisting of an Apple-style Web Dashboard and a Manifest V3 Chrome Browser Extension.

The system leverages a hybrid detection pipeline combining a locally trained Machine Learning Stylistic Classifier and deep semantic Gemini AI Fact-Checking with real-time Google Search grounding.

---

🚀 Key Features

• Apple Pro Obsidian Theme: A modern, high-contrast, pure pitch-black dashboard with responsive card highlights and smooth scroll-snapping layout animations.
• Hybrid Classification Pipeline:
  1. Stylistic Classification (ML): Evaluates linguistic structures, sensationalism, and stylistic syntax using TF-IDF and Logistic Regression.
  2. Semantic Verification (AI): Extracts specific claims from articles and fact-checks them against active Google Search indexes using Gemini.
• Social Media Scraping Fallback: Standard scraping blocks on social media sites (Facebook, X, etc.) are bypassed by using Gemini Google Search Grounding to fetch verified context on URLs.
• Chrome Extension: Scan the active webpage with a single click, fetching immediate credibility verdicts, clickbait scores, and claim-level details directly from your browser toolbar.
• SQLite Database Logging: Scan records are stored locally, letting you browse, search, filter, and audit detailed historical reports.

---

📊 Datasets & Machine Learning Model

The Local ML Classifier is trained on a merged corpus of public datasets loaded from Hugging Face:

1. GonzaloA/fake_news (24,353 records)
2. ErfanMoosaviMonazzah/fake-news-detection-dataset-English (30,000 records)

Data Normalization

The labels in both datasets were normalized into a single unified format:

• 0 = Reliable / Real
• 1 = Misleading / Fake

Model Performance

• Vectorization: TF-IDF Vectorizer with unigrams and bigrams.
• Model: Logistic Regression.
• Test Accuracy: 98.19% accuracy achieved on the merged 54,000+ record test split.
• Model Files: Saved locally in backend/data/ for instantaneous inference inside FastAPI endpoints:
  • tfidf_vectorizer.joblib
  • fake_news_model.joblib

---

🛠️ Project Structure

├── backend/            # Python FastAPI backend
│   ├── app/            # Application logic (routes, database, analyzer, ML model)
│   ├── data/           # Trained models (.joblib) and local SQLite database (.db)
│   ├── train.py        # ML training script utilizing Hugging Face datasets
│   └── requirements.txt# Python backend dependencies
├── frontend/           # Web Dashboard files
│   ├── index.html      # Dashboard markup with scroll-snap welcome page
│   ├── style.css       # Obsidian Dark CSS design system
│   └── app.js          # Chart rendering and API integrations
├── extension/          # Manifest V3 Chrome Extension
│   ├── manifest.json   # Chrome Extension configuration
│   ├── popup.html      # Extension layout
│   ├── popup.css       # Obsidian-themed styling
│   └── popup.js        # Tab scraper and backend connector
└── README.md           # Documentation

---

⚙️ Setup & Installation

Step 1: Run the Backend API Server

1. Navigate to the backend directory:

   cd backend

2. Create and activate a Python virtual environment:

   python -m venv .venv
   # On Windows:
   .venv\Scripts\activate
   # On macOS/Linux:
   source .venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Start the development server:

   python -m uvicorn app.main:app --host 127.0.0.1 --port 8000

   The backend will boot up at http://127.0.0.1:8000.

Step 2: Open the Web Dashboard

Open your browser and navigate to http://localhost:8000 (or http://127.0.0.1:8000) to view the dashboard. Scroll down to enter and explore.

Step 3: Install the Chrome Extension

1. Open Google Chrome and go to chrome://extensions/.
2. Enable Developer mode (toggle in top-right).
3. Click Load unpacked (top-left).
4. Select the extension/ folder in this project repository.
5. The extension is now loaded! Pin it to your toolbar to start scanning articles.

Step 4: Configure Gemini AI Fact-Checking

1. Go to the Settings tab in the Web Dashboard.
2. Enter your Google Gemini API Key (get one free from Google AI Studio).
3. Click Save API Key. The system will dynamically switch from heuristics mode to active claim fact-checking!