Chatbot RAG Demo

This Demo Bot is a chatbot application designed to improve LLM-based question answering and reduce hallucinations by integrating web scraping, FAQ processing, and text-based embeddings for advanced question-answering capabilities. It utilizes Python, FastAPI, Flask, HuggingFace embeddings, and ChromaDB for semantic search and data persistence. Currently it is applied to a TA Chatbot for a highly technical course in computer science.

Table of Contents

• Features
• Repo Structure
• Setup and Installation
• Usage
• Technical Details

Features

• Frontend and Backend Integration:

  • A user-friendly Flask-based frontend (chatbot_app.py).
  • A FastAPI backend (app.py) for intelligent processing and answer generation.

• Semantic Search with ChromaDB:

  • Employs 🤗 HuggingFace all-MiniLM-L6-v2 embeddings for semantic matching.
  • Uses ChromaDB for efficient vector storage and retrieval.

• Web Scraping and Data Processing:

  • Scrapes course-related content and FAQs with Scrapy (scrape_website.py).
  • Parses and organizes raw web content and FAQs into structured formats (parse_website.py).

• Context-Aware Responses:

  • Combines FAQs and content snippets to generate detailed, contextually rich answers.

• Persistent Storage:

  • Raw scraped data is saved for reusability.
  • Embeddings and processed data are persisted using ChromaDB.

Repo Structure

.
├── chroma/                           # Directory for ChromaDB data persistence
├── raw_webcraw_data/                 # Directory for raw scraped HTML and images
├── raw_webcraw_data_faq_processed/   # Processed FAQ data from web scraping
├── templates/                        
│   └── chatbot.html                  # Chatbot frontend HTML
├── .gitignore                        # Files to ignore in Git
├── app.py                            # Main bot file
├── chatbot_app.py                    # Flask frontend
├── FAQBot.ipynb                      # Test notebook for the chatbot (deprecated)
├── parse_website.py                  # Web parsing logic for FAQ and course data
├── README_crawler.md
├── README_demobot.md
├── README.md
├── requirements.txt                  # Python dependencies
├── scrape_website.py                 # Scrapy-based web scraper

Technical Details

Semantic Search and ChromaDB

• HuggingFace Embeddings: The HuggingFace all-MiniLM-L6-v2 model is used to embed text into dense vector representations.
• ChromaDB: These embeddings are stored and managed using ChromaDB, enabling efficient semantic search across both course content (chroma/db) and FAQs (chroma/db_faq).

Web Scraping and Data Preparation

• Scrapy for Data Collection:
  • scrape_website.py crawls the specified domain and collects raw HTML pages and images, stored under raw_webcraw_data/.
• HTML Parsing and FAQ Extraction:
  • parse_website.py processes raw HTML into structured sections, extracting FAQ question-answer pairs. Processed FAQs are saved in raw_webcraw_data_faq_processed/.

Question Answering Workflow

1. User Input:
   • Users submit questions via the chatbot frontend.
2. FAQ Matching:
   • The backend searches for matching questions in the FAQ database.
3. Content Search:
   • Relevant course snippets are retrieved from ChromaDB using semantic search.
4. Answer Generation:
   • OpenAI GPT-4o-mini (Or Gemini 1.5 Flash) generates context-aware answers by combining FAQ and course snippets.

Persistent Storage

• Raw Data:
  • Scraped HTML and images are stored in raw_webcraw_data/.
  • Processed FAQ data is saved in raw_webcraw_data_faq_processed/.
• Embedding Databases:
  • chroma/db: Course content embeddings.
  • chroma/db_faq: FAQ embeddings.

Setup and Installation

Prerequisites

1. Python 3.8 or higher.
2. Clone the repository:

   git clone https://github.com/jsz-05/Chatbot-RAG.git
   cd Chatbot-RAG

3. Install dependencies from requirements.txt:

   pip install -r requirements.txt

4. Set up environment variables: Create a .env file in the project root with the following content:

   OPENAI_API_KEY=<your_api_key>
   OPENAI_ORGANIZATION=<your_organization_id>
   

1. Backend API

Start the FastAPI backend:

uvicorn app:app --host 0.0.0.0 --reload

Access the API documentation at:

http://127.0.0.1:8000/docs

2. Frontend Chatbot

Run the Flask chatbot frontend:

python3 chatbot_app.py

Access the chatbot on:

http://127.0.0.1:5000/