VeriGraph-AI 🛡️

VeriGraph-AI is an advanced agentic AI system built with LangGraph and LangChain that orchestrates a team of specialized AI workers to answer complex user queries. The system features an intelligent supervisor that routes tasks, a Retrieval-Augmented Generation (RAG) engine, real-time web scraping capabilities, and a robust validation loop to ensure high-quality responses.

🚀 Features

• Supervisor-Worker Architecture: A central supervisor node analyzes user intent and routes queries to the most appropriate worker.
• Specialized Workers:
  • 🤖 LLM Node: Handles general conversational queries and logic using Groq.
  • 📚 RAG Node: Retrieves precise information from a local knowledge base (e.g., US Economy data) using ChromaDB and HuggingFace embeddings.
  • 🌐 Web Scraper Node: Fetches real-time information from the internet using the Tavily Search API.
• ✅ Self-Correction & Validation: A dedicated validator node reviews all outputs. If an answer is deemed invalid or insufficient, it rejects the result and triggers a retry or alternative routing.
• State Management: Built on LangGraph's StateGraph to maintain robust conversation state and handle complex cyclic workflows.

🏗️ Architecture

The system operates as a directed graph where:

1. Supervisor receives the user query.
2. Router directs the flow to LLM, RAG, or Web Scraper based on the supervisor's decision.
3. Worker executes the task and generates a response.
4. Validator evaluates the response against the original query.
   • If Valid: The workflow ends and returns the result.
   • If Invalid: The workflow loops back to the Supervisor to attempt a different approach.

graph TD
    Start((Start)) --> Supervisor(Supervisor)
    
    Supervisor -.-> LLM(LLM)
    Supervisor -.-> RAG(RAG)
    Supervisor -.-> WebScraper(Web Scraper)
    
    LLM --> Validator(Validator)
    RAG --> Validator(Validator)
    WebScraper --> Validator(Validator)
    
    Validator -.->|Invalid| Supervisor
    Validator -.->|Valid| End((End))
    
    %% Styling with explicit black text for high contrast
    style Start fill:#d1c4e9,stroke:#512da8,stroke-width:2px,color:#000000
    style Supervisor fill:#bbdefb,stroke:#0d47a1,stroke-width:2px,color:#000000
    style LLM fill:#c8e6c9,stroke:#1b5e20,stroke-width:2px,color:#000000
    style RAG fill:#c8e6c9,stroke:#1b5e20,stroke-width:2px,color:#000000
    style WebScraper fill:#c8e6c9,stroke:#1b5e20,stroke-width:2px,color:#000000
    style Validator fill:#ffccbc,stroke:#bf360c,stroke-width:2px,color:#000000
    style End fill:#d1c4e9,stroke:#512da8,stroke-width:2px,color:#000000

Loading

🛠️ Prerequisites

• Python 3.12+
• Jupyter Notebook (to run the core logic)
• API Keys for:
  • Groq (LLM inference)
  • Tavily (Web Search)
  • HuggingFace (Embeddings)
  • LangSmith (Optional, for tracing)

📦 Installation

1. Clone the Repository

   git clone https://github.com/yourusername/verigraph-ai.git
   cd verigraph-ai

2. Install Dependencies It is recommended to use a virtual environment.

   # Using pip
   pip install -r requirements.txt
   
   # Or if using uv/poetry (based on pyproject.toml)
   pip install .

3. Set Up Environment Variables Create a .env file in the code/ directory or project root with the following keys:

   GROQ_API_KEY=your_groq_api_key
   TAVILY_API_KEY=your_tavily_api_key
   HUGGINGFACE_API_KEY=your_huggingface_api_key
   LANGCHAIN_API_KEY=your_langchain_api_key
   LANGCHAIN_TRACING_V2=true
   LANGCHAIN_PROJECT=VeriGraph-AI

🚀 Usage

1. Navigate to the Code Directory

   cd code

2. Launch Jupyter Notebook

   jupyter notebook

3. Run the Workflow Open index_updated.ipynb. Run all cells to initialize the graph.

   You can test the agent with queries like:

   • "What is the capital of France?" (LLM)
   • "Tell me about the US GDP in 2024." (RAG - requires data/usa.txt)
   • "What is the current stock price of Apple?" (Web Scraper)

   # Example invocation in the notebook
   app.invoke({"messages": ["What is the current inflation rate in the US?"]})

📂 Project Structure

verigraph-ai/
├── code/
│   └── index_updated.ipynb   # Main workflow definition and logic
├── data/
│   └── usa.txt               # Knowledge base for RAG
├── pyproject.toml            # Project dependencies
├── README.md                 # Project documentation
└── .env                      # Environment variables (not committed)

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

This project is open-source and available under the MIT License.