https://github.com/hurkanugur/rag-based-chatbot
An intelligent retrieval-augmented (RAG) PDF Chatbot: upload a PDF and ask questions, answers generated from its content using local embeddings and a lightweight model.
https://github.com/hurkanugur/rag-based-chatbot
deep-learning gradio gradio-interface gradio-ui hugging-face huggingface huggingface-transformers python pytorch rag rag-chatbot rag-pdf transformer transformer-architecture transformers
Last synced: 28 days ago
JSON representation
An intelligent retrieval-augmented (RAG) PDF Chatbot: upload a PDF and ask questions, answers generated from its content using local embeddings and a lightweight model.
- Host: GitHub
- URL: https://github.com/hurkanugur/rag-based-chatbot
- Owner: hurkanugur
- License: mit
- Created: 2025-10-25T12:33:20.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2025-10-25T17:31:58.000Z (7 months ago)
- Last Synced: 2025-10-25T18:17:36.587Z (7 months ago)
- Topics: deep-learning, gradio, gradio-interface, gradio-ui, hugging-face, huggingface, huggingface-transformers, python, pytorch, rag, rag-chatbot, rag-pdf, transformer, transformer-architecture, transformers
- Language: Python
- Homepage:
- Size: 6.84 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# 🧠 RAG-Based ChatBot (Offline)
## 📖 Overview
This project implements a **Retrieval-Augmented Generation (RAG) chatbot** that answers questions based on **user-provided PDF documents**.
- 📄 **Dynamic knowledge base** — user uploads PDFs which are split into paragraphs
- ⚙️ **Embeddings using Sentence Transformers** — converts text into vectors
- 🧱 **FAISS-based retrieval** — efficiently finds the most relevant paragraphs
- 🤖 **Text generation using Flan-T5** — generates answers using retrieved context
- 💻 **Local inference** — runs on **CPU, Apple MPS, or NVIDIA CUDA GPU**
- 🖥️ **Gradio UI** — web interface for uploading PDFs and asking questions
- 🛠️ **Lazy model loading** — model loads only on first query for faster startup
---
## 🖼️ Application Screenshot
Below is a preview of the **Gradio Interface** used for real-time classification:
---
## 🧩 Libraries
- **PyTorch** – model and inference
- **Transformers** – Flan-T5 text generation
- **sentence-transformers** – embeddings
- **faiss** – vector retrieval
- **numpy** – numerical operations
- **PyPDF2** – PDF parsing
- **gradio** – UI interface
---
## ⚙️ Requirements
- Python **3.13+**
- Recommended editor: **VS Code**
---
## 📦 Installation
- Clone the repository
```bash
git clone https://github.com/hurkanugur/RAG-Based-ChatBot.git
```
- Navigate to the `RAG-Based-ChatBot` directory
```bash
cd RAG-Based-ChatBot
```
- Install dependencies
```bash
pip install -r requirements.txt
```
---
## 🔧 Setup Python Environment in VS Code
1. `View → Command Palette → Python: Create Environment`
2. Choose **Venv** and your **Python version**
3. Select **requirements.txt** to install dependencies
4. Click **OK**
---
## 📂 Project Structure
```bash
assets/
└── app_screenshot.png # Application screenshot
src/
├── config.py # Paths, model settings...
├── device_manager.py # Selects and manages compute device
├── embeddings.py # Local embedding model
├── retriever.py # FAISS-based retrieval
├── generator.py # Text generation model
├── pdf_loader.py # PDF parsing
└── ui.py # Gradio UI
main/
└── main.py # Entry point for chatbot
requirements.txt # Python dependencies
```
---
## 📂 System Architecture
```bash
User PDF Upload
↓
PDF Loader (split into paragraphs)
↓
Sentence Transformer Embeddings
↓
FAISS Retriever (top-k relevant paragraphs)
↓
Flan-T5 Generator (with retrieved context)
↓
Answer displayed in Gradio UI
```
---
## 📂 Run the Application
Navigate to the project directory:
```bash
cd RAG-Based-ChatBot
```
Run the app:
```bash
python -m main.main
```
or
```bash
python3 -m main.main
```