Agentic RAG Chatbot with Streamlit and LangGraph

This project is an advanced Retrieval-Augmented Generation (RAG) chatbot featuring an agentic workflow powered by LangGraph. Users can upload a PDF and ask questions about its content via a Streamlit interface. If the answer isn’t found in the document, the agent can search the web automatically.

---

✨ Features

• PDF Processing: Upload and extract text from any PDF.
• Agentic Workflow: Stateful, multi-step agent using LangGraph for reasoning and decision-making.
• Confidence-Based Routing: Answers from the PDF if confident, otherwise falls back to web search.
• Web Search Fallback: Uses Tavily AI for external queries.
• Document Summarization: summarize command generates a concise summary.
• Interactive UI: Simple web interface built with Streamlit.

---

🔑 API Keys Required

• Google Gemini API key
• Tavily AI API key

---

⚙️ How It Works

The application uses a stateful graph (LangGraph) to define agent workflow paths. Each step is a node, and transitions are directed by edges.

Visual Workflow

graph TD
    A(Start: User uploads PDF & asks question) --> B[initialize_models];
    B --> C[load_pdf_text];
    C --> D[chunk_text];
    D --> E[create_chroma_index];
    E --> F[retrieve];

    F --> G{route_after_retrieval};
    G -- "Query is 'summarize'" --> H[summarize];
    G -- "Normal question" --> I[evaluate];

    I --> J{route_after_evaluation};
    J -- "Confidence ≥ 0.6" --> K[generate];
    J -- "Confidence < 0.6" --> L[web_search];

    L --> K;

    K --> M(End: Display Answer);
    H --> M;

    classDef start-end fill:#28a745,stroke:#333,stroke-width:2px,color:#fff;
    classDef process fill:#007bff,stroke:#333,stroke-width:2px,color:#fff;
    classDef decision fill:#ffc107,stroke:#333,stroke-width:2px,color:#000;
    classDef special-path fill:#fd7e14,stroke:#333,stroke-width:2px,color:#fff;

    class A,M start-end;
    class B,C,D,E,F,I,K,L,H process;
    class G,J decision;

Loading

🚀 Getting Started

1. Prerequisites

• Python 3.10+ (Recommended Python 3.14.2)
• Poetry or venv

2. Installation

Clone the repository and install the dependencies:

# Create and activate virtual environment
python -m venv .venv
.\.venv\Scripts\Activate.ps1  # Windows
source .venv/bin/activate     # Unix/macOS

# Install dependencies
pip install -r requirements.txt

3. Configuration

Create a .env file in the root directory (using .env.example as a template):

GEMINI_API_KEY=your_gemini_api_key
TAVILY_API_KEY=your_tavily_api_key

4. Usage

Run the Streamlit application:

streamlit run AgenticRAG_using_LangGraph.py

📁 Project Structure

• AgenticRAG_using_LangGraph.py: Main application logic and Streamlit UI.
• data/: Sample PDF documents for testing.
• .env.example: Template for environment variables.
• requirements.txt: Project dependencies.