Fact-Checking News Aggregator

A multi-agent system that automatically crawls news sources, extracts factual claims, verifies them against reliable sources, and publishes verified results to a Jekyll website.

🎯 Project Description

This project implements an automated fact-checking pipeline using multiple AI agents that work together to:

1. Crawl news sources for articles
2. Extract factual claims from articles
3. Verify claims against reliable sources
4. Publish verified results to a Jekyll website

The system uses the A2A (Agent-to-Agent) protocol for inter-agent communication and MCP (Model Context Protocol) for external tool integration, creating a robust and scalable fact-checking infrastructure.

🏗️ Architecture

Multi-Agent System

• Orchestrator Agent: Coordinates the entire pipeline and manages agent communication
• Crawler Agent: Fetches news articles from various sources
• Extractor Agent: Extracts factual claims from articles using LLM analysis
• Fact Checker Agent: Verifies claims against reliable sources
• Publisher Agent: Publishes verified results to the Jekyll website

Communication Protocol

• A2A Protocol: Enables asynchronous communication between agents
• MCP Integration: Provides external tools for claim extraction, verification, and publishing

🛠️ Tools & Technologies

Core Technologies

• Python 3.11+: Main programming language
• Jekyll: Static site generator for the fact-checking website
• A2A Protocol: Agent-to-agent communication framework
• MCP (Model Context Protocol): External tool integration

AI/ML Components

• Large Language Models: For claim extraction and verification
• Semantic Analysis: For understanding and processing news content
• Fact Verification: Against reliable databases and sources

Web Technologies

• Jekyll: Static site generation
• Markdown: Content formatting
• CSS/HTML: Website styling and structure

🚀 Quick Start

Prerequisites

• Python 3.11 or higher
• Ruby and Jekyll (for website generation)
• Git

Installation

1. Clone the repository

   git clone <repository-url>
   cd Fact_Checking_News_Aggregator

2. Set up Python environment

   # Using uv (recommended)
   uv sync
   
   # Or using pip
   pip install -r requirements.txt

3. Install Jekyll (for website generation)

   gem install jekyll jekyll-feed

Running the System

1. Run the complete pipeline

   python run_all_agents.py

   This will:

   • Start the mcp server
   • Initialize all agents

2. Start the Orchestrator

   python orchestrator.py

   This will:

   • Start the Orchestrator Agent.

3. Start the Orchestrator Client

   python orchestrator_client.py

   This will:

   • Start the client interface
   • Type 'start' to run the pipeline and 'end' to stop the flow.

4. Serve the Jekyll website

   cd jekyll_site
   jekyll serve --host 0.0.0.0 --port 4000

   Visit http://localhost:4000 to see the fact-checked results.

🔧 MCP-A2A Integration

MCP Server Tools

The MCP server provides three main tools that agents can call:

1. extract_claims: Extracts factual claims from news articles

   • Input: Article text
   • Output: JSON array of claims

2. verify_claim: Verifies a claim against reliable sources

   • Input: Claim statement
   • Output: Verification result (True/False) and source

3. generate_jekyll_post: Creates Jekyll blog posts for verified claims

   • Input: Statement, verification status, source
   • Output: Generated Markdown file with proper Jekyll format

A2A Protocol Integration

• Asynchronous Communication: Agents communicate asynchronously using handle_message_async
• Message Routing: Orchestrator routes messages between agents
• Error Handling: Robust error handling for failed agent communications
• State Management: Each agent maintains its own state and processing logic

Agent Workflow

1. Orchestrator → Crawler: Request news articles
2. Crawler → Orchestrator: Return article URLs and content
3. Orchestrator → Extractor: Send articles for claim extraction
4. Extractor → MCP: Call extract_claims tool
5. Extractor → Orchestrator: Return extracted claims
6. Orchestrator → Fact Checker: Send claims for verification
7. Fact Checker → MCP: Call verify_claim tool
8. Fact Checker → Orchestrator: Return verification results
9. Orchestrator → Publisher: Send verified claims for publishing
10. Publisher → MCP: Call generate_jekyll_post tool

📁 Project Structure

Fact_Checking_News_Aggregator/
├── agents/                    # Agent implementations
│   ├── crawler_agent/        # News crawling agent
│   ├── extractor_agent/      # Claim extraction agent
│   ├── fact_checker_agent/   # Fact verification agent
│   └── publisher_agent/      # Publishing agent
├── jekyll_site/              # Jekyll website
│   ├── _posts/              # Generated fact-check posts
│   ├── _layouts/            # Jekyll layouts
│   ├── assets/              # CSS and styling
│   └── _config.yml          # Jekyll configuration
├── mcp_server.py            # MCP server with external tools
├── orchestrator.py          # Main orchestrator agent
├── run_all_agents.py        # Pipeline execution script
└── requirements.txt         # Python dependencies

🔍 How It Works

1. News Crawling

The crawler agent fetches news articles from configured sources and extracts relevant content.

2. Claim Extraction

The extractor agent uses LLM analysis to identify factual claims within articles, calling the MCP extract_claims tool.

3. Fact Verification

The fact checker agent verifies each claim against reliable sources using the MCP verify_claim tool.

4. Publishing

The publisher agent generates Jekyll blog posts for verified claims using the MCP generate_jekyll_post tool.

5. Website Generation

Jekyll automatically builds the website with the new fact-checked posts, making them available at the configured URL.

🎨 Customization

Adding News Sources

Modify the crawler agent configuration to add new news sources.

Custom Verification Sources

Update the fact checker agent to use different verification databases or APIs.

Website Styling

Customize the Jekyll theme and styling in the jekyll_site/assets/ directory.

Agent Behavior

Modify individual agent configurations to adjust their behavior and processing logic.

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

Note: This system is designed for educational and research purposes. Always verify fact-checking results independently for critical applications.