Food Recall Report Agent

A multi-agent system for analyzing and reporting on food recall news in the United States.

Overview

This system utilizes LangChain and Google's Gemini LLM to:

1. Collect food recall data from FDA and USDA websites
2. Extract key information from recall announcements
3. Analyze potential economic impact
4. Generate comprehensive weekly reports

System Architecture

The system consists of four specialized agents:

• Data Collection Agent: Fetches recall data from FDA and USDA websites
• Information Extraction Agent: Processes raw data to identify key details
• Economic Impact Agent: Estimates financial consequences of recalls
• Reporting Agent: Generates weekly reports ranking recalls by severity and impact

These agents are coordinated by an Orchestrator that manages the workflow.

Data Sources

• FDA: Web scraping of the FDA recalls page
• USDA: Web scraping of the USDA FSIS recalls page

Installation

1. Clone this repository
2. Install dependencies:

pip install -r requirements.txt

3. Create a .env file with your API keys:

GOOGLE_API_KEY=your_gemini_api_key
TAVILY_API_KEY=tavily_key
FIRECRAWL_API_KEY=firecrawl_key

Usage

Run the orchestrator to execute the complete pipeline:

python main.py
python main.py --days 14

Or run individual steps:

python main.py --step collect
python main.py --step extract
python main.py --step analyze
python main.py --step report

Project Structure

.
├── data/                  # Storage for collected and processed data
│   ├── raw/               # Raw data from FDA API and USDA website
│   ├── processed/         # Structured data after information extraction
│   └── analyzed/          # Data with economic impact analysis
├── reports/               # Output directory for generated reports
├── src/
│   ├── agents/            # Implementation of specialized agents
│   ├── models/            # Data models and schemas
│   ├── utils/             # Helper functions and utilities
│   └── orchestrator.py    # Workflow coordinator
├── main.py                # Main entry point
├── test_fda_api.py        # FDA API test script
└── README.md