Agentic RAG-based QA System with LangChain 🦜️🔗

A sophisticated question-answering system implementing Retrieval-Augmented Generation (RAG) with agentic decision-making capabilities using LangChain and LangGraph.

Key Features ✨

• Agentic Workflow: Stateful decision-making process using LangGraph
• Dynamic Document Retrieval: ChromaDB vector store with configurable chunking
• Relevance Grading: LLM-powered document relevance assessment
• Self-Improving Queries: Automatic query rewriting for better retrieval
• Modular Architecture: Easily extendable components for customization
• Multi-Model Support: Configurable OpenAI models for different tasks
• Evaluation Ready: Pre-configured test cases with answer verification

Installation ⚙️

1. Clone repository:

git clone https://github.com/dkleptsov/agentic_rag.git
cd agentic_rag

2. Install dependencies:

pip install -r requirements.txt

3. Set up environment variables:

echo "OPENAI_API_KEY=your_api_key_here" > .env

Configuration 🛠️

Modify config.py to customize system behavior:

# Data handling
DATABASE_DIR = "vector_db"            # ChromaDB storage location
DATA_FILE = "data/anna_karenina.txt"  # Source document

# Model configurations
GRADER_MODEL = "gpt-4o"               # Document relevance grading
REWRITE_MODEL="gpt-4o-2024-08-06"     # Query rewriting
GENERATE_MODEL = "gpt-4o-mini"        # Answer generation
AGENT_MODEL = "gpt-4-turbo"           # Decision-making agent
OPENAI_TEMPERATURE = 0                # Temperature of the model answers

# Text processing
CHUNK_SIZE = 1000                     # Document chunk size
CHUNK_OVERLAP = 200                   # Chunk overlap

# Evaluation
QUESTION_ANSWER_PAIRS = [             # Test questions with expected answers
    ("What is Anna Karenina's opening line?", 
     "Happy families are all alike..."),
    # Add more QA pairs
]

Usage 🚀

from agentic_rag import ask_question

question = "What is the novel's perspective on family dynamics?"
answer = ask_question(question)
print(f"Answer: {answer}")

Example output:

Question: What is Anna Karenina's opening line?
RAG answer: The famous opening line of Leo Tolstoy's Anna Karenina is:
"Happy families are all alike; every unhappy family is unhappy in its own way."
Correct answer: Happy families are all alike...

System Architecture 🏛️

graph TD
    A[User Question] --> B{Agent Decision}
    B -->|Retrieve| C[Document Retrieval]
    C --> D{Relevance Grader}
    D -->|Relevant| E[Generate Answer]
    D -->|Irrelevant| F[Rewrite Question]
    F --> B
    E --> G[Final Answer]

Loading

Core Components:

1. Intelligent Agent (LLM)

• Makes retrieval decisions
• Manages conversation state
• Orchestrates workflow

2. Document Retriever

• ChromaDB vector store
• Recursive text splitting
• Semantic search

3. Relevance Grader

• Binary document assessment
• Prevents irrelevant context usage
• Configurable threshold

4. Query Rewriter

• Improves search effectiveness
• Semantic query transformation
• Feedback loop integration

5. Answer Generator

• RAG-powered responses
• Multi-model support
• Streamable output

Structure of the project 📁

├── chroma_db                       # Directory for ChromaDB vector store
├── data                            # Directory for input data
│   └── anna-karenina.txt           # Text file containing the source document
├── images                          # Directory for visual assets
│   └── Agentic_RAG.png             # Diagram or screenshot of the system architecture
├── agentic_rag.py                  # Main script for the Agentic RAG-based QA system
├── config.py                       # Configuration file for model and system settings
├── README.md                       # Project documentation and overview
└── requirements.txt                # List of Python dependencies for the project

Contributing 🤝

Contributions are welcome! Please follow these steps:

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

Disclaimer ⚠️

This system uses OpenAI language models. Please ensure:

• You comply with OpenAI's usage policies
• Sensitive data is properly handled
• API costs are monitored