Simple RAG

This example demonstrates how to build a simple RAG system that loads documents from files and answers questions about their content.

Overview

This example shows:

Loading documents from text files
Basic question answering
Source citation and confidence scoring
Error handling and validation

Prerequisites

Python 3.8+
RecoAgent installed (pip install recoagent)
OpenAI API key
Sample text files for the knowledge base

Code Implementation

import os
import glob
from pathlib import Path
from dotenv import load_dotenv
from recoagent import RecoAgent

# Load environment variables
load_dotenv()

class SimpleRAG:
    def __init__(self):
        """Initialize the RAG system."""
        self.agent = RecoAgent(
            llm_provider="openai",
            llm_model="gpt-3.5-turbo",
            embedding_model="text-embedding-ada-002",
            chunk_size=500,
            chunk_overlap=50
        )
        
    def load_documents_from_files(self, directory: str):
        """Load all text files from a directory."""
        documents = []
        file_paths = []
        
        # Find all .txt files in directory
        txt_files = glob.glob(os.path.join(directory, "*.txt"))
        
        for file_path in txt_files:
            try:
                with open(file_path, 'r', encoding='utf-8') as file:
                    content = file.read().strip()
                    if content:
                        documents.append(content)
                        file_paths.append(file_path)
                        print(f"✅ Loaded: {os.path.basename(file_path)}")
            except Exception as e:
                print(f"❌ Error loading {file_path}: {e}")
        
        # Add documents to agent
        if documents:
            self.agent.add_documents(documents)
            print(f"\n📚 Total documents loaded: {len(documents)}")
            return file_paths
        else:
            print("❌ No documents found or loaded")
            return []
    
    def ask_question(self, question: str):
        """Ask a question and get a detailed response."""
        print(f"\n🤔 Question: {question}")
        
        try:
            response = self.agent.ask(question)
            
            print(f"💡 Answer: {response.answer}")
            print(f"🎯 Confidence: {response.confidence:.2f}")
            print(f"📚 Sources: {len(response.sources)} documents")
            
            # Show source details
            if response.sources:
                print("\n📖 Source Documents:")
                for i, source in enumerate(response.sources, 1):
                    # Truncate long sources
                    source_preview = source[:100] + "..." if len(source) > 100 else source
                    print(f"  {i}. {source_preview}")
            
            return response
            
        except Exception as e:
            print(f"❌ Error processing question: {e}")
            return None
    
    def batch_questions(self, questions: list):
        """Ask multiple questions and show results."""
        print(f"\n📋 Processing {len(questions)} questions...")
        
        results = []
        for i, question in enumerate(questions, 1):
            print(f"\n--- Question {i}/{len(questions)} ---")
            response = self.ask_question(question)
            if response:
                results.append({
                    'question': question,
                    'answer': response.answer,
                    'confidence': response.confidence
                })
        
        return results

def create_sample_files():
    """Create sample text files for demonstration."""
    sample_dir = "sample_docs"
    os.makedirs(sample_dir, exist_ok=True)
    
    # Sample documents
    docs = {
        "company_info.txt": """
        RecoHut is a leading provider of enterprise AI solutions.
        Founded in 2023, we specialize in RAG (Retrieval-Augmented Generation) systems.
        Our flagship product, RecoAgent, helps organizations build intelligent knowledge management systems.
        We serve clients across healthcare, finance, legal, and technology sectors.
        """,
        
        "product_features.txt": """
        RecoAgent Key Features:
        - Hybrid retrieval combining BM25 and vector search
        - Built-in evaluation with RAGAS metrics
        - Safety guardrails using NVIDIA NeMo Guardrails
        - Multiple vector store support (OpenSearch, Azure AI Search, Vertex AI)
        - LangSmith integration for observability
        - FastAPI-based REST API for easy integration
        """,
        
        "technical_details.txt": """
        Technical Architecture:
        RecoAgent is built using Python and follows a modular architecture.
        Core components include:
        - packages.agents: Agent orchestration using LangGraph
        - packages.rag: Retrieval, reranking, and evaluation
        - packages.observability: Monitoring, tracing, and metrics
        
        The system supports both local and cloud deployments.
        Minimum requirements: Python 3.8+, 4GB RAM, 10GB storage.
        """
    }
    
    # Write sample files
    for filename, content in docs.items():
        filepath = os.path.join(sample_dir, filename)
        with open(filepath, 'w', encoding='utf-8') as file:
            file.write(content.strip())
    
    print(f"✅ Created sample documents in {sample_dir}/")
    return sample_dir

def main():
    """Main function to run the simple RAG example."""
    print("🚀 Simple RAG Example")
    print("=" * 40)
    
    # Create sample files if they don't exist
    sample_dir = create_sample_files()
    
    # Initialize RAG system
    rag = SimpleRAG()
    
    # Load documents
    print(f"\n📁 Loading documents from {sample_dir}/...")
    file_paths = rag.load_documents_from_files(sample_dir)
    
    if not file_paths:
        print("❌ No documents loaded. Exiting.")
        return
    
    # Interactive Q&A session
    print("\n" + "=" * 40)
    print("💬 Interactive Q&A Session")
    print("=" * 40)
    print("Type 'quit' to exit, 'batch' for batch questions")
    
    while True:
        user_input = input("\n❓ Ask a question: ").strip()
        
        if user_input.lower() == 'quit':
            break
        elif user_input.lower() == 'batch':
            # Run batch questions
            batch_questions = [
                "What is RecoHut?",
                "What are the key features of RecoAgent?",
                "What is the technical architecture?",
                "What are the minimum requirements?"
            ]
            rag.batch_questions(batch_questions)
        elif user_input:
            rag.ask_question(user_input)
    
    print("\n👋 Thanks for using Simple RAG!")

if __name__ == "__main__":
    main()

Running the Example

1. Setup

# Create project directory
mkdir simple-rag-example
cd simple-rag-example

# Install dependencies
pip install recoagent python-dotenv

# Create .env file
echo "OPENAI_API_KEY=your_api_key_here" > .env

2. Run the Example

python simple_rag.py

3. Expected Output

🚀 Simple RAG Example
========================================

✅ Created sample documents in sample_docs/

📁 Loading documents from sample_docs/...
✅ Loaded: company_info.txt
✅ Loaded: product_features.txt
✅ Loaded: technical_details.txt

📚 Total documents loaded: 3

========================================
💬 Interactive Q&A Session
========================================
Type 'quit' to exit, 'batch' for batch questions

❓ Ask a question: What is RecoHut?

🤔 Question: What is RecoHut?
💡 Answer: RecoHut is a leading provider of enterprise AI solutions, founded in 2023. We specialize in RAG (Retrieval-Augmented Generation) systems and serve clients across healthcare, finance, legal, and technology sectors.
🎯 Confidence: 0.92
📚 Sources: 1 documents

📖 Source Documents:
  1. RecoHut is a leading provider of enterprise AI solutions. Founded in 2023, we specialize in RAG...

❓ Ask a question: batch

📋 Processing 4 questions...

--- Question 1/4 ---

🤔 Question: What is RecoHut?
💡 Answer: RecoHut is a leading provider of enterprise AI solutions, founded in 2023...
🎯 Confidence: 0.92
📚 Sources: 1 documents

--- Question 2/4 ---

🤔 Question: What are the key features of RecoAgent?
💡 Answer: RecoAgent's key features include hybrid retrieval combining BM25 and vector search, built-in evaluation with RAGAS metrics, safety guardrails, multiple vector store support, LangSmith integration, and FastAPI-based REST API.
🎯 Confidence: 0.89
📚 Sources: 1 documents

❓ Ask a question: quit

👋 Thanks for using Simple RAG!

Understanding the Code

Key Components

Document Loading: Loads all .txt files from a directory
Error Handling: Gracefully handles file loading errors
Batch Processing: Can process multiple questions at once
Source Citation: Shows which documents were used for answers
Confidence Scoring: Provides confidence levels for responses

Customization Options

# Custom chunking strategy
rag = SimpleRAG()
rag.agent = RecoAgent(
    chunk_size=300,      # Smaller chunks
    chunk_overlap=100,   # More overlap
    chunking_strategy="semantic"  # Semantic chunking
)

# Different LLM models
rag.agent = RecoAgent(
    llm_model="gpt-4",           # Use GPT-4
    temperature=0.1,             # Lower temperature for consistency
    max_tokens=500               # Limit response length
)

Troubleshooting

Common Issues

No documents loaded:

# Check file permissions and encoding
ls -la sample_docs/
file sample_docs/*.txt

API key issues:

# Verify API key is set
echo $OPENAI_API_KEY

# Test API key
python -c "import openai; openai.api_key='$OPENAI_API_KEY'; print('API key valid')"

Low confidence scores:

Ensure documents contain relevant information
Try rephrasing questions
Check document quality and formatting

Next Steps

This example provides a foundation for more advanced RAG systems. You can extend it by:

Adding more file formats (PDF, Word, HTML)
Implementing document preprocessing (cleaning, formatting)
Adding metadata tracking (file names, timestamps)
Creating a web interface for easier interaction
Adding evaluation metrics to measure performance

Minimal Usage - Even simpler setup
Vector Search - Focus on retrieval
Basic Agent - Add agent capabilities

Ready for more? Check out the Vector Search Example to learn about different retrieval strategies!

Overview​

Prerequisites​

Code Implementation​

Running the Example​

1. Setup​

2. Run the Example​

3. Expected Output​

Understanding the Code​

Key Components​

Customization Options​

Troubleshooting​

Common Issues​

Next Steps​

Related Examples​