/ragURL

This project aim to retrieve the information from the provided URL

Primary LanguagePython

URLinsights: A RAG-based URL Content Analysis System

Executive Summary

URLinsights is a Retrieval Augmented Generation (RAG) system that enables intelligent question-answering based on URL content. The system combines modern NLP techniques, vector databases, and large language models to provide accurate, context-aware responses while maintaining source traceability.

Table of Contents

Approach

  1. Environment Setup

    • Created dedicated Python environment for URLinsights
    • Configured necessary dependencies and libraries

  2. Core Functionality Implementation

    • Utilized UnstructuredURLLoader (LangChain Community) for URL content extraction
    • Implemented text chunking (size: 1024, overlap: 100) for optimal embedding creation
    • Integrated sentence-transformers all-mpnet-base-v2 model for embeddings
    • Implemented FAISS for vector storage
    • Integrated Ollama and Groq for answer retrieval

  3. User Interface Development

    • Built Streamlit-based frontend with features including:
      • User authentication
      • URL management (add/remove/select)
      • Multi-URL information retrieval
      • Chat history management

  4. Deployment

    • Deployed on AWS EC2
    • Extended EBS storage for dependencies
    • Implemented RAG system features (source tracking, context management)

Technology Stack

Category Technologies
Backend Python
NLP Framework Langchain, Ollama, Groq
LLM Models llama3
Vector Store FAISS
Frontend Streamlit
Cloud AWS EC2, AWS EBS

Challenges & Solutions

  1. Environment Setup

    • Challenge: Limited internet speed
    • Solution: Optimized dependency management and installation process

  2. Embedding Generation

    • Challenge: Slow performance with OllamaEmbeddings
    • Solution: Migrated to HuggingfaceEmbeddings for improved speed

  3. Text Generation

    • Challenge: High latency with Ollama (≈2 minutes)
    • Solution: Integrated Groq for near-instant responses

  4. Deployment

    • Challenges:
      • Limited AWS free tier resources
      • Storage constraints
      • CPU limitations for embeddings
    • Solutions:
      • Optimized resource usage
      • Implemented efficient storage management

  5. Context Management

    • Challenge: Token limits in extended conversations
    • Solution: Implemented context windowing and management

Potential Improvements

  1. Scalability & Performance

    • Integrate AWS Bedrock for improved scalability
    • Implement OpenSearch for parallel search capabilities
    • Optimize chunking parameters and strategy

  2. Model Enhancements

    • Explore advanced embedding models
    • Evaluate alternative Language Models
    • Integrate Vision Language Models for image analysis

  3. Architecture

    • Implement better chunking strategies
    • Optimize parameter configurations
    • Add support for multimedia content analysis

Installation

# Clone the repository
git clone https://github.com/Harshit-tech9/ragURL.git

# Navigate to project directory
cd URLinsights

# Install dependencies
pip install -r requirements.txt

# Set up environment variables
cp .env.example .env
# Edit .env with your configurations

Usage

# Start the Streamlit application
streamlit run app.py

# Access the application
# Open browser and navigate to: http://localhost:8501

Conclusion

URLinsights demonstrates the potential of RAG systems for URL content analysis. The current implementation provides robust functionality for content extraction and analysis, while the proposed improvements outline a path for enhanced capabilities, performance, and scalability.

UI Screens