Deployed - https://med-bot-ephi.onrender.com

🧠 MediBot — AI-Powered Chatbot for Q&A over Medical PDFs

MediBot is an AI-powered medical chatbot that leverages state-of-the-art language models and vector search to answer user queries based on a curated set of medical PDF documents. It uses Streamlit for the user interface, LangChain for LLM orchestration, and FAISS for efficient vector search.

---

📚 Table of Contents

• Features
• Project Structure
• Setup Instructions
  • 1. Clone the Repository
  • 2. Install Dependencies
  • 3. Prepare Environment Variables
  • 4. Prepare Data
  • 5. Create Vector Store
  • 6. Run the Chatbot
• How It Works
• Customization
• Troubleshooting
• Disclaimer
• License

---

✨ Features

• 📄 Medical PDF Knowledge Base: Answers are grounded in the provided medical PDFs.
• 🤖 Modern LLM Integration: Uses Groq LLaMA-3.1-8B-Instant for high-quality responses.
• 🔎 Semantic Search: FAISS-powered vector search for relevant context retrieval.
• 🖥️ Streamlit UI: Simple, interactive chat interface.
• 🧩 Custom Prompting: Easily modify the prompt template for different behaviors.

---

📁 Project Structure

.
├── .env
├── connect_memory_with_llm.py
├── create_memory_for_llm.py
├── app.py
├── Pipfile
├── Pipfile.lock
├── data/
│   ├── clinical_medicine_ashok_chandra.pdf
│   ├── ...
│   └── vectorstore/
│       └── db_faiss/
└── ...

• MediBot.py: Main Streamlit app for the chatbot.
• create_memory_for_llm.py: Script to process PDFs and build the FAISS vector store.
• connect_memory_with_llm.py: Script to test the QA chain in the terminal.
• data/: Folder containing all source PDFs.
• vectorstore/: Stores the FAISS vector database.

---

⚙️ Setup Instructions

1. Clone the Repository

git clone <your-repo-url>
cd medical_chatBot

2. Install Dependencies

It's recommended to use pipenv (or you can use pip):

pipenv install
pipenv shell

Or, using pip:

pip install -r requirements.txt

3. Prepare Environment Variables

Create a .env file in the project root with your Groq API key:

GROQ_API_KEY=your_groq_api_key_here

Alternatively, set the environment variable in your shell:

export GROQ_API_KEY=your_groq_api_key_here

4. Prepare Data

• Place all your medical PDF files inside the data/ directory.

5. Create Vector Store

Run the following script to process PDFs and build the FAISS vector store:

python create_memory_for_llm.py

This will:

• Load all PDFs from data/
• Split them into text chunks
• Generate embeddings using the sentence-transformers/all-MiniLM-L6-v2 model
• Store the embeddings in vectorstore/db_faiss

6. Run the Chatbot

Start the Streamlit app:

streamlit run MediBot.py

Open the provided local URL in your browser to interact with MediBot.

---

🔍 How It Works

1. Document Loading: All PDFs in data/ are loaded and split into manageable text chunks.
2. Embedding Generation: Each chunk is embedded using the sentence-transformers/all-MiniLM-L6-v2 model.
3. Vector Store: Embeddings are stored in a FAISS vector database for efficient similarity search.
4. Query Handling: When a user asks a question, the most relevant chunks are retrieved from FAISS.
5. LLM Response: The context and question are sent to the Groq LLaMA-3.1-8B-Instant model via the Groq API.
6. Answer Display: The answer is shown in the Streamlit chat interface, along with the source documents if needed.

---

🛠️ Customization

• Prompt Template:
  Modify the prompt in MediBot.py or connect_memory_with_llm.py to change the chatbot's behavior.
• Model Selection:
  Change the MODEL_NAME variable to use a different Groq LLaMA model.
• Chunk Size:
  Adjust chunk_size and chunk_overlap in create_memory_for_llm.py for different granularity.

---

🐞 Troubleshooting

• FAISS Not Found:
  Ensure you have run create_memory_for_llm.py before starting the chatbot.
• API Key Issues:
  Double-check your Groq API key in .env or your environment.
• Dependency Errors:
  Make sure all dependencies are installed. Use pipenv or pip as described above.
• CUDA/CPU Issues:
  If running on a machine without GPU, ensure models are set to run on CPU.

---

⚠️ Disclaimer

This chatbot provides AI-generated information and should not be considered a substitute for professional medical advice. Always consult a qualified doctor for medical concerns.

---

📄 License

This project is for educational and research purposes only. Please check the licenses of the models and datasets you use.