Yoni.T
Back to Projects
AI/Data/MLFeb 2024 - Jul 20243 Min Read

Implementing an AI Chatbot for Enhanced Information Access with LLM and RAG

Development of an AI-based chatbot using the RAG method at Catur Insan Cendekia University to enhance fast and contextual information access.

Implementing an AI Chatbot for Enhanced Information Access with LLM and RAG

🤖 AI-Based Chatbot System with Retrieval-Augmented Generation (RAG)

🌐 Overview

In today’s digital era, quick and accurate information access is crucial, especially in education. To address this need, I developed an AI-based chatbot system for Catur Insan Cendekia University using the Retrieval-Augmented Generation (RAG) method. This system provides fast, relevant, and contextual information access for students and the public.

❓ What is Retrieval-Augmented Generation (RAG)?

RAG is an innovative approach that combines two main components:

  1. 🗂️ Indexing and Retrieval
  2. 📝 Generation

This method enables the system to:

  • 🔍 Search and retrieve relevant information from a database.
  • 💬 Process and generate contextual, informative responses.

⚙️ How It Works

1. Indexing

  • 📄 Documents are structured hierarchically into “parent chunks” and “child chunks”.
  • 🧠 Each chunk is transformed into vectors using embedding models like Sentence-BERT to improve semantic accuracy.

2. Retrieval

  • 🔎 Utilizes a Parent Document Retriever to locate the most relevant data based on user queries.
  • 🎯 Pinpoints the exact child chunk while providing the parent chunk as context for a comprehensive response.

3. Augmented Information

  • ✨ The retrieved data is enhanced with:
    • 🗣️ Language model instructions.
    • 🗃️ Vector database context.
    • 💡 User query inputs.

4. Generation

  • 🚀 The enhanced information is processed using a large language model (LLM) such as Gemini AI to generate coherent and informative responses.

🛠️ Tools Used

To build this RAG-based chatbot, the following tools and technologies were utilized:

  • 🧰 LangChain: Framework for building applications powered by LLMs.
  • 🤗 HuggingFace: Pre-trained NLP models for embedding and text processing.
  • 🗄️ ChromaDB: Vector database for storing embeddings.
  • 🌌 Gemini AI: Large language model for generating responses.
  • 📦 Unstructured IO: Document processing and structuring.
  • 🌐 Django: Backend framework for building the API.
  • ⚡ Redis: Fast data storage for caching.
  • 🐝 Celery: Task queue for asynchronous processing.
  • 🔧 Other Python libraries to support the implementation.

✨ Features

  • ⚡ Fast and Relevant Responses: Provides accurate and contextual answers to user queries.
  • 📚 Information Hierarchy: Structures documents to ensure efficient retrieval.
  • 🎓 Educational Focus: Tailored to meet the needs of students and the academic community.

🖼️ Screenshots

Chat Interface

Chat Interface 1 Chat Interface 2

Admin Interface

Admin Interface 1 Admin Interface 2

🌟 Impact

By applying the RAG method, this chatbot significantly improves the accessibility and efficiency of information services in educational settings. It offers a scalable solution for modern institutions aiming to enhance their digital services.

📂 Repository

This project demonstrates the practical application of AI-based chatbots and RAG techniques in real-world scenarios. For further details, please visit the GitHub repository:

🔗 Chatbot Admin System

📄 Published Article

The related article has been published and can be accessed at:
🔗 Seminar Nasional IAII - SISFOTEK Journal

💡 Skills Demonstrated

  • 🤖 Large Language Models (LLM)
  • 🧠 Retrieval-Augmented Generation (RAG)
  • 📝 Natural Language Processing (NLP)
  • 🧰 LangChain Framework
  • 🔍 Information Retrieval
  • 🌐 Backend Development with Django