Agentic RAG System with AutoGen

An advanced Retrieval-Augmented Generation (RAG) system built with Microsoft's AutoGen framework, featuring autonomous AI agents that collaborate to answer questions based on your documents.

🚀 Features

• Multi-Agent Architecture: Orchestrator, Retriever, and Analyzer agents working together
• AutoGen Framework: Leverages Microsoft AutoGen for agent collaboration
• OpenAI Integration: Uses ChatOpenAI (GPT-4o-mini) and OpenAI Embeddings (text-embedding-3-small)
• Vector Search: Semantic search using LangChain Chroma with OpenAI embeddings
• Document Processing: Supports PDF, DOCX, and TXT file formats
• Flexible Configuration: YAML-based configuration for easy customization
• Interactive CLI: User-friendly command-line interface for querying

📋 System Architecture

The system consists of three main agents:

1. Orchestrator Agent: Coordinates the overall workflow and manages agent interactions
2. Document Retriever Agent: Searches the vector database for relevant documents
3. Content Analyzer Agent: Analyzes retrieved documents and generates comprehensive answers

Workflow

User Query → Orchestrator → Retriever → Vector Search → Retrieved Docs
                    ↓                                          ↓
              Final Answer ← Analyzer ← Document Analysis ←────┘

🛠️ Installation

Prerequisites

• Python 3.8 or higher
• pip package manager

Setup Steps

1. Clone the repository (or navigate to project directory):

   cd /workspaces/sample.ai

2. Install dependencies:

   pip install -r requirements.txt

3. Configure environment variables:

   cp .env.example .env

   Edit .env and add your OpenAI API key:

   OPENAI_API_KEY=your_openai_api_key_here
   

📁 Project Structure

sample.ai/
├── src/
│   ├── __init__.py
│   ├── document_processor.py    # Document processing and chunking
│   ├── vectorstore.py            # Vector store management
│   └── agents.py                 # AutoGen agent definitions
├── data/
│   └── documents/                # Place your documents here
│       ├── ai_introduction.txt
│       └── rag_explained.txt
├── config.yaml                   # Configuration file
├── main.py                       # Main application script
├── ingest_documents.py           # Document ingestion utility
├── requirements.txt              # Python dependencies
└── README.md                     # This file

🚀 Usage

Step 1: Add Your Documents

Place your documents in the data/documents/ directory. Supported formats:

• PDF files (.pdf)
• Word documents (.docx)
• Text files (.txt)

Step 2: Ingest Documents

Process and index your documents:

python ingest_documents.py

To clear existing documents and start fresh:

python ingest_documents.py --clear

Step 3: Run the Agentic RAG System

Start the interactive query interface:

python main.py

Example Interaction

================================================================================
Agentic RAG System with AutoGen
================================================================================
Initializing vector store...
Vector store already contains 47 documents.

Initializing Agentic RAG system...
System initialized successfully!

================================================================================
You can now ask questions about your documents.
Type 'exit' or 'quit' to end the session.
================================================================================

Your question: What is RAG and how does it work?

Processing your query...

[Agent interactions occur here...]

Answer: RAG (Retrieval-Augmented Generation) is a technique that combines 
large language models with external knowledge retrieval. It works by...

⚙️ Configuration

Edit config.yaml to customize:

LLM Settings

llm:
  model: "gpt-4o-mini"  # or "gpt-4o", "gpt-4", etc.
  temperature: 0.7
  max_tokens: 2000

Embedding Configuration

embedding:
  model: "text-embedding-3-small"  # or "text-embedding-3-large"
  chunk_size: 1000
  chunk_overlap: 200

Vector Store Settings

vectorstore:
  type: "chromadb"
  persist_directory: "./vectorstore"
  collection_name: "agentic_rag_collection"

🎯 Key Components

Document Processor

• Handles multiple file formats (PDF, DOCX, TXT)
• Implements intelligent text chunking with LangChain
• Preserves document metadata

Vector Store Manager

• LangChain Chroma integration for vector storage
• OpenAI embeddings (text-embedding-3-small)
• Semantic similarity search with score ranking

AutoGen Agents

• RetrieverAgent: Searches and retrieves relevant documents
• AnalyzerAgent: Analyzes content using ChatOpenAI
• OrchestratorAgent: Coordinates multi-agent workflow
• Function calling for document retrieval

🔧 Advanced Usage

Custom Document Processing

from src.document_processor import DocumentProcessor

processor = DocumentProcessor(chunk_size=500, chunk_overlap=100)
documents = processor.process_directory("path/to/documents")

Direct Vector Store Access

from src.vectorstore import VectorStoreManager

vectorstore = VectorStoreManager()
results = vectorstore.search("your query", n_results=10)

Programmatic Querying

from src.agents import AgenticRAGSystem

rag_system = AgenticRAGSystem(vectorstore, llm_config)
answer = rag_system.query("What is machine learning?")

🐛 Troubleshooting

Issue: "OPENAI_API_KEY not found"

Solution: Make sure you've created a .env file with your API key.

Issue: No documents found

Solution: Run python ingest_documents.py to process documents first.

Issue: Import errors

Solution: Ensure all dependencies are installed: pip install -r requirements.txt

📚 Dependencies

• pyautogen: Multi-agent framework
• langchain & langchain-openai: LLM and embeddings integration
• langchain-chroma: Vector database with LangChain
• chromadb: Vector storage backend
• openai: OpenAI API client
• PyPDF2: PDF processing
• python-docx: Word document processing

🤝 Contributing

Contributions are welcome! Areas for improvement:

• Additional document format support
• More sophisticated chunking strategies
• Agent memory and conversation history
• Web UI interface
• Multiple LLM provider support

📄 License

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

🙏 Acknowledgments

• Microsoft AutoGen framework
• ChromaDB for vector storage
• OpenAI for LLM capabilities
• The open-source AI community

📞 Support

For issues, questions, or contributions, please open an issue on the repository.

---

Built with ❤️ using AutoGen and modern AI technologies