RAG-Based Chatbot for Smart Customer Support Documents

This project implements a Retrieval-Augmented Generation (RAG) based chatbot to intelligently answer questions using a PDF document containing both text and images. It uses local embedding models and the Groq API for high-performance, cost-free generation, avoiding any paid APIs.

💡 Features

• Extracts both text (using PyMuPDF) and image data from PDF using OCR (Tesseract).
• Converts PDF into vector embeddings using all-MiniLM-L6-v2.
• Stores the embeddings in a FAISS vector database.
• Uses the Groq API (LLaMA3) for generating context-based answers.
• Runs locally in Visual Studio / VS Code using Python virtual environment.
• Simple test script (test.py) to query the bot.
• Clean architecture for maintainability and extension.

---

📁 Project Structure

• .gitignore – Files/folders ignored by Git
• app.py – 🚀 Streamlit or UI interface
• CSR MODULES.pdf – 📄 Input PDF file (text + image content)
• images/ – 🌅 Auto-extracted images from PDF for OCR
• ingest_pdf.py – ⚙️ PDF parser, OCR, and embedding logic
• rag_chatbot.py – 🤖 Core RAG chatbot pipeline
• requirements.txt – 📦 Python dependencies
• test.py – 🧪 CLI script to interact with the chatbot
• vectorstore/ – 🧠 Local FAISS vector database
• README.md – 📘 You're reading it now!

🧠 Architecture Diagram

• General RAG Workflow

• Implemented RAG Chatbot Architecture

🛠️ Setup Instructions

1. Clone the Repository

• git clone https://github.com/balahariharasudhan/RAG-Based-Chatbot-for-Smart-Customer-Support-Documents.git

• Change the directory to the cloned repo by using -> cd RAG-Based-Chatbot-for-Smart-Customer-Support-Documents

2. Create and activate a Virtual Environment (To work separately without disturbing the existing versions)

• python -m venv venv

• .\venv\Scripts\activate # For Windows

3. Install Requirements

• pip install -r requirements.txt

4. Configure Environment Variables (For ensuring security)

• Create a .env file and store your Groq API (Free):

• It should be in the form of: GROQ_API_KEY=your_groq_api_key

5. 📄 Ingest PDF

• You can either use the built-in PDF or upload your personal company documents as a PDF and chat with the bot

• Run the following to process your PDF and store the vector embeddings: python ingest_pdf.py

After successful running, it will:

• Extracts text.

• Extracts images and runs OCR using Tesseract.

• Embeds content using Sentence Transformers.

• Stores embeddings in FAISS DB (vectorstore/).

6. 🤖 Query Your Chatbot

• Use the test.py script to ask questions by: python test.py

7. 🌐 Launch Streamlit UI

• After successful completion, make sure to run the UI by using: streamlit run app.py

---

Sample interaction:

Q: What are the key objectives of CSR? A: [Answer from context]

💬 Model Used

1. Embeddings: all-MiniLM-L6-v2 from Sentence Transformers (free, local).

2. LLM: llama3-8b-8192 via Groq API (free and fast inference).

📸 Screenshots

🔹 Sample Query and Response in terminal

🔹 Streamlit UI

---

✅ Key Learnings

• Used LangChain (0.3.27) with new modular packages like:

-> langchain-community

-> langchain-core

-> langchain-huggingface

• Enabled FAISS with allow_dangerous_deserialization=True for local use.

• OCR setup using pytesseract + pdf2image + Tesseract executable path config.

• Dealt with deprecated imports and updated LangChain compatibility manually.

🧠 Future Scope

• Expand to real-time document search (external knowledge integration).

🙋‍♂️ Author

• Balahariharasudhan T

• GitHub: https://github.com/balahariharasudhan

📜 License

• This project is for educational and research purposes only.