Fact-Check Agent

A multi-agent fact-checking system that verifies user claims using Wikipedia as a data source. Built with Google's Agents Development Kit (ADK) and LangChain.

Overview

This project implements an automated fact-checking workflow with two specialized agents:

1. Fact Researcher: Searches Wikipedia to gather evidence relevant to a claim
2. Fact Checker: Evaluates the claim against research data and provides a verdict

The system uses a sequential workflow where the researcher gathers evidence first, and then the fact-checker evaluates the claim based on that evidence.

Features

• Multi-agent architecture using Google ADK
• Wikipedia integration for evidence gathering
• Automatic claim verification with detailed reasoning
• Structured verdicts: TRUE, FALSE, PARTIALLY TRUE, or UNVERIFIED
• Google Cloud logging integration
• Environment variable configuration

Prerequisites

• Python 3.x
• Google Cloud credentials (for authentication)
• .env file with MODEL environment variable set

Installation

1. Clone the repository:

git clone <repository-url>
cd fact-check-agent

2. Install dependencies:

pip install -r requirements.txt

3. Set up environment variables: Create a .env file in the project root:

MODEL=<your-model-name>

4. Configure Google Cloud authentication: Set up your Google Cloud credentials for authentication.

Usage

Import and use the fact-checking workflow:

from agent import root_agent, fact_check_workflow

# Use the root agent to interact with users
# or directly call the fact_check_workflow for automated fact-checking

The workflow will:

1. Accept a claim from the user
2. Research the claim using Wikipedia
3. Evaluate the claim against the gathered evidence
4. Return a verdict with detailed reasoning

Project Structure

fact-check-agent/
├── __init__.py           # Package initialization
├── agent.py              # Main agent definitions and workflow
├── requirements.txt      # Python dependencies
└── README.md             # This file

Dependencies

• google-adk==1.14.0 - Google's Agents Development Kit
• langchain-community==0.3.27 - LangChain community tools
• wikipedia==1.4.0 - Wikipedia API client

How It Works

1. The Greeter Agent provides an entry point and collects the user's claim
2. The Fact Researcher Agent uses Wikipedia tools to search for relevant information
3. The Fact Checker Agent analyzes the research data and provides a final verdict with supporting evidence