Medical RAG Chatbot

A medical question-answering chatbot leveraging Retrieval-Augmented Generation (RAG) for safe, context-aware responses using medical documents. Built with Streamlit, Hugging Face, FAISS, and PyTorch.

---

Features

• Retrieval-Augmented Generation (RAG) for enhanced answer accuracy.
• Embedding medical documents using SentenceTransformers.
• Context-aware and memory-enabled conversation.
• Simple, interactive Streamlit web interface.
• Prioritizes user safety; does not answer out-of-context medical questions.

---

Tech Stack

• Python 3.8+
• Streamlit for the UI
• PyTorch for deep learning
• Sentence-Transformers for text embeddings
• FAISS for vector search
• LangChain for prompt construction
• transformers for LLM pipelines
• Other: pandas, numpy, nltk, PyPDF2

---

Getting Started

Prerequisites

• Python 3.8 or above
• pip (Python package manager)
• Medical data files (data/index.faiss, data/meta.csv)

Installation

1. Clone the repository:

   git clone https://github.com/Thamannahafeez/medical-rag-chatbot.git
   cd medical-rag-chatbot

2. Install dependencies:

   pip install -r requirements.txt

3. Place your FAISS index and metadata files in the data/ folder.

Usage

Start the Streamlit app:

streamlit run app.py

The app will be available at http://localhost:8501 by default.

---

Project Structure

File/Folder	Description
app.py	Main Streamlit application file
requirements.txt	Python dependencies
data/index.faiss	FAISS embedding index
data/meta.csv	Chunk metadata for retrieval

---

Data

• You can supply a FAISS index file containing embeddings and a corresponding metadata CSV in the data/ directory. This repo contains proprietary medical data embedded from a PDF version of a medical textbook named "Gupte-The-Short-Textbook-of-Pediatrics-11th-Ed-2009" .

---

Limitations and Disclaimer

• The chatbot answers only using context from the uploaded documents. If the answer is not known or context is insufficient, it will refuse to respond or direct the user to a clinician.
• Do not use for clinical diagnosis or decision-making.

---

Acknowledgments

• Based on open-source RAG chatbot architectures and inspired by similar medical FAQ bots
• Uses components from HuggingFace, SentenceTransformers, and FAISS.

---