🤖 Agentic RAG using CrewAI

A powerful Retrieval-Augmented Generation (RAG) system built with CrewAI that intelligently searches through documents and falls back to web search when needed. Features local LLM support with deep-seek-r1 or llama 3.2!

🌟 Features

• 📚 Document-based search with RAG capabilities
• 🌐 Automatic fallback to web search
• 🤖 Local LLM support (deep-seek-r1 or llama 3.2)
• 🔄 Seamless integration with CrewAI
• 💨 Fast and efficient document processing
• 🎯 Precise answer synthesis

🔄 System Flow

Below is the detailed flow diagram of how the system processes queries and generates responses:

graph TD
    A[Start] --> B[Initialize Streamlit App]
    B --> C[Load LLM Model]
    C --> D[Initialize Session State]
    
    D --> E{PDF Uploaded?}
    E -->|Yes| F[Create DocumentSearchTool]
    E -->|No| G[Wait for PDF Upload]
    
    F --> H[Index PDF Document]
    H --> I[Create Crew]
    
    I --> J[Create Retriever Agent]
    I --> K[Create Response Synthesizer Agent]
    
    J --> L[Add Tools to Retriever Agent]
    L --> L1[PDF Search Tool]
    L --> L2[Web Search Tool]
    
    K --> M[Configure Response Agent]
    
    J & K --> N[Create Tasks]
    N --> N1[Retrieval Task]
    N --> N2[Response Task]
    
    N --> O[User Enters Query]
    
    O --> P[Process Query]
    P --> Q[Show User Message]
    Q --> R[Crew Kickoff]
    
    R --> S[Sequential Processing]
    S --> T1[Retriever Agent Searches]
    T1 --> T2[Response Agent Synthesizes]
    
    T2 --> U[Stream Response]
    U --> V[Update Chat History]
    
    V --> W[Wait for Next Query]
    W --> O

Loading

🚀 Prerequisites

Before running the application, ensure you have:

1. API Keys:

   • FireCrawl API or SEPER API key for web search capabilities
   • LLM API key (if required for your chosen model)

2. Python Environment:

   • Python 3.11 or later
   • Conda (recommended for environment management)

💻 Installation

1. Create and Activate Environment:

   conda create -n env_crewai python==3.12 -y
   conda activate env_crewai

2. Install Dependencies:

   # Install package management tools
   uv lock
   uv sync
   
   # Install required packages
   pip install crewai crewai-tools markitdown qdrant-client fastembed

🎮 Running the Application

Choose your preferred LLM model:

• For deep-seek-r1:

  streamlit run app_deep_seek.py

• For llama 3.2:

  streamlit run app_llama3.2.py

🛠️ System Architecture

The system consists of two main agents:

1. Retriever Agent:

   • Handles document searching
   • Manages web search fallback
   • Uses both PDF and web search tools

2. Response Synthesizer Agent:

   • Processes retrieved information
   • Generates coherent responses
   • Ensures context relevance

📚 Usage Examples

1. Document Search:

   • Upload your PDF document
   • Enter your query
   • Receive contextual answers from the document

2. Web Search Fallback:

   • System automatically detects when document search isn't sufficient
   • Seamlessly switches to web search
   • Combines information from multiple sources

🤝 Contributing

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

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• CrewAI for the amazing framework
• The open-source community for various tools and libraries used in this project

---

Made with ❤️ by Rahul Samant