ENSAH Assistant (ENSAH-RAG)

A conversational AI assistant powered by RAG (Retrieval-Augmented Generation) technology to provide information about ENSA Al-Hoceima (École Nationale des Sciences Appliquées d'Al-Hoceima).

🏗️ Architecture

🌟 Features

• Interactive chat interface built with Streamlit
• RAG-based responses using LangChain and Google Generative AI
• Vector storage with Qdrant for efficient information retrieval
• Comprehensive web scraping system for ENSAH website data
• Multi-language support
• Real-time information about:
  • Academic programs and departments
  • News and announcements
  • Scientific activities
  • Faculty members
  • Administrative information
  • General information about ENSAH
• Persistent chat history using MongoDB
  • Maintains conversation context across sessions
  • Stores user interactions and responses
  • Enables personalized experiences
• Session Management
  • Unique session IDs for each user
  • Local storage integration for session persistence
  • Stateful conversations across page refreshes

🛠️ Technology Stack

• Frontend: Streamlit
• Backend: FastAPI , LangServe
• Vector Database: QdrantDB (Cloud storage)
• LLM Integration: LangChain, Google Generative AI
• Web Scraping: BeautifulSoup4
• Data Storage: MongoDB (for chat history)
• Monitoring: LangSmith
• Containerization: Docker

📁 Project Structure

ENSAH-RAG/
├── DataENG/          # English language data
├── DataFR/           # French language data (original data)
├── Image/            # Core RAG and API handlers
├── QdrantDB/         # Vector database storing scripts
├── RawData/          # Raw scraped data
├── Scrapers/         # Web scraping modules
├── assets/          # Static assets
├── app.py           # Streamlit frontend
├── docker-compose.yml
├── requirements.txt
└── README.md

## 🚀 Getting Started

### Prerequisites

- Python 3.8+
- Docker and Docker Compose
- Google AI API key
- QdrantDB API key
- LangSmith API key

### Installation

1. Clone the repository:
```bash
git clone https://github.com/ELhadratiOth/ENSAH-ChatBot-RAG-APP
cd ENSAH-RAG

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Set up environment variables: rename a sample.env file to .env and fill it with:

GOOGLE_API_KEY=GOOGLE_API_KEY
QDRANT_API_KEY=QDRANT_API_KEY
QDRANT_URL=QDRANT_URL
LANGSMITH_API_KEY=LANGSMITH_API_KEY

5. Start the services using Docker Compose:

docker-compose up -d

6. Run the application:

streamlit run app.py

🔧 Components

Web Scrapers

Located in the Scrapers/ directory, includes modules for scraping:

• News and announcements
• Faculty information
• Academic programs
• Administrative information
• Scientific activities
• General information

RAG System

The Image/ directory contains the core RAG implementation:

• Vector store management
• Chat history handling
• API endpoints
• Prompt templates
• Response generation

Frontend

The Streamlit interface (app.py) provides:

• Chat interface
• Source attribution
• Session management
• Responsive design

📸 User Interface

Here are some screenshots of the ENSAH Assistant interface:

Chat History and Session Management

The system implements sophisticated chat history and session handling:

MongoDB Chat History

• Persistent storage of chat messages in MongoDB
• Each conversation is linked to a unique session ID
• Enables context-aware responses by maintaining conversation history
• Custom LastNMessageHistory implementation for efficient history management

Session Management System

• Generates unique session IDs for each user
• Integrates with local storage for session persistence
• Maintains user context across page refreshes
• Seamless conversation continuity between visits

This robust history and session management enables:

• Context-aware conversations
• Personalized user experiences
• Consistent chat flow across sessions
• Improved response relevancy through historical context

🤝 Contributing

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