HighlightFactCheck

Overview

HighlightFactCheck is an advanced AI-powered fact-checking service designed to provide instant, comprehensive analysis of textual content. Our system leverages state-of-the-art language models and multi-source verification to deliver thorough, nuanced fact-checks quickly and efficiently.

Key Features

• AI-Powered Analysis: Utilizes advanced GPT models for nuanced understanding
• Multi-Source Verification: Checks Google Fact Check Tools, custom web searches, and recent news
• Comprehensive Results: Provides ratings, severity assessments, and detailed explanations
• Continuous Learning: Stores fact-checks for faster future retrievals and improved accuracy
• Rewrite Suggestions: Offers AI-powered suggestions to improve statement accuracy
• Secure Authentication: Utilizes Auth0 for robust user authentication
• Fast Processing: Delivers fact-check results within minutes
• Wide Content Support: Fact-checks articles, social media posts, and more
• Historical Data: Access to previously fact-checked content for quick reference

Project Structure

.
├── app.py                 # Main Flask application
├── fact_checker.py        # Core fact-checking logic
├── db
│   ├── facts_db.py        # Database operations for fact storage
│   └── user_db.py         # User database operations
├── config
│   ├── config.py          # Configuration settings
│   └── api_config.py      # API-specific configuration
├── tools
│   └── logger.py          # Logging utility
├── templates
│   ├── index.html         # Main landing page
│   ├── members.html       # Members area template
│   └── search.html        # Search functionality template
├── static
│   └── images
│       ├── logo-trans.gif # Transparent logo
│       └── chrome-flow.gif # Chrome extension demo
├── requirements.txt       # Project dependencies
└── README.md              # This file

Setup and Installation

1. Clone the repository:

   git clone https://github.com/your-username/HighlightFactCheck.git
   cd HighlightFactCheck
   

2. Create a virtual environment and activate it:

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`
   

3. Install the required dependencies:

   pip install -r requirements.txt
   

4. Set up environment variables: Create a .env file in the root directory and add the following variables:

   GOOGLE_API_KEY=your_google_api_key
   OPENAI_API_KEY=your_openai_api_key
   AUTH0_CLIENT_ID=your_auth0_client_id
   AUTH0_CLIENT_SECRET=your_auth0_client_secret
   AUTH0_DOMAIN=your_auth0_domain
   APP_SECRET_KEY=your_app_secret_key
   STRIPE_SECRET_KEY=your_stripe_secret_key
   STRIPE_WEBHOOK_SECRET=your_stripe_webhook_secret
   MYSQL_HOST=your_mysql_host
   MYSQL_USER=your_mysql_user
   MYSQL_PASSWORD=your_mysql_password
   MYSQL_DATABASE=your_mysql_database
   MYSQL_PORT=your_mysql_port
   

5. Run the application:

   flask run (from folder containing app.py)
   

Core Components

1. Flask Application (app.py)

The main Flask application handles routing, user authentication, and integrates all components. Key routes include:

• /: Landing page
• /login: Auth0 login
• /callback: Auth0 callback
• /members: Members area
• /check: Fact-checking endpoint for paid users
• /check-free: Rate-limited fact-checking for free users
• /webhook: Stripe webhook for subscription management

2. Fact Checker (fact_checker.py)

The core logic for fact-checking, including:

• Sentence tokenization
• Google Fact Check Tools API integration
• Custom web searches (commented out the call for google cse because of cost)
• News article analysis
• AI-powered claim relevance determination
• Comprehensive fact-check generation

3. Database Operations

• facts_db.py: Manages fact storage using LanceDB
• user_db.py: Handles user data and subscription status in MySQL

4. Authentication and Payment

• Auth0 integration for secure user authentication
• Stripe integration for subscription management

TODO List - Add them here as you need to.

• .com and similar tld cause the form the split a question up resulting in false answers that are not true.

• Develop API documentation for potential future public release

• Implement user feedback mechanism for continuous improvement

• Add meme fact checker for image fact checking pipeline. Extract text from pictures, use this as the input. Make a free limited version and unlimited members routes and all them to the html pages.

Notes for Founders

• The current pricing model is set at $99.99/month for unlimited fact-checks. This can be much less because we ditched google CSE
• The SIMILARITY_THRESHOLD in api_config.py is currently set to 0.95. We might need to adjust this based on user feedback and system performance.
• The fact-checking system currently uses GPT-4.
• The Chrome extension is not live to the public yet. This should be a priority for the next development sprint.