Agentic Supply Chain Risk Intelligence System

A data-driven, multi-agent AI system for analyzing supply chain risks using statistical modeling and LLM-based investigation.

Features

• Statistical Risk Scoring: Automatically flags risky suppliers using Z-Scores, Percentiles (90th), and Composite Risk Scores.
• Multi-Agent AI:
  • Auditor Agent: Analyzes internal data for anomalies.
  • Investigator Agent: Searches the web for external disruptions (strikes, weather, politics).
  • Manager Agent: Generates strategic action plans (Retain/Monitor/Replace).
• Interactive Dashboard: Built with Streamlit and Plotly for deep data exploration.
• Predictive Modeling: Uses Random Forest to estimate the probability of future disruptions.
• Automated EDA: Instant data quality reports and statistical summaries.

Installation

1. Clone the repository:

   git clone <your-repository-url>
   cd agentic-supply-chain

2. Install dependencies:

   pip install -r requirements.txt

   Note: Ensure you have plotly, crewai, and scikit-learn installed. If you encounter issues with streamlit not being found, follow the usage steps below.

3. Set up API Keys: This app requires a Google Gemini API Key (for the LLM) and a Serper API Key (for Google Search). You can enter them in the app sidebar.

Usage

1. Run the Application: Since streamlit might not be in your global PATH, use the module invocation:

   python -m streamlit run src/app.py

   Alternatively, you can run the provided batch script if available:

   src/run_app.bat

2. Analyze Data:

   • Upload CSV: Upload your own supplier data.
   • Use Demo Data: Toggle the checkbox to load built-in synthetic data for testing.

3. Kickoff Agents:

   • Scroll down to the "Agentic Investigation" section.
   • Click Kickoff Agents.
   • The system will sequentially audit suppliers, check external news, and generate a final report.

Project Structure

├── src/
│   ├── app.py               # Modular entry point (Run this)
│   ├── agents/              # CrewAI Agent definitions
│   ├── analysis/            # Risk scoring & Feature engineering logic
│   ├── models/              # Predictive models
│   ├── visualization/       # Plotly chart generation
│   └── full_app.py          # Legacy monolithic version
├── data/
│   └── sample_suppliers.csv # Sample dataset
├── requirements.txt         # Python dependencies
└── README.md                # This file

Testing

Run the test suite to verify the logic:

python -m pytest

License

MIT License