💊 MedRec - AI-Powered Medicine Recommendation & Info Chatbot

MedRec is an advanced, production-ready AI chatbot designed to provide instant, accurate, and reliable information about medicines. Built using a robust Retrieval-Augmented Generation (RAG) architecture, it leverages a massive dataset of medicine information, storing it in a high-performance vector database to deliver context-aware answers using Google's Gemini LLM.

⚠️ Disclaimer: MedRec is designed for informational and educational purposes only. It is not a substitute for professional medical advice, diagnosis, or treatment. Always seek the advice of your physician or other qualified health provider with any questions you may have regarding a medical condition.

---

✨ Key Features

• 🧠 Advanced RAG Pipeline: Combines the reasoning power of Gemini 2.5 with a highly structured medical database to prevent hallucination and ensure accurate responses.
• ⚡ Serverless & Lightweight: Completely relies on cloud APIs (Hugging Face & Gemini) for ML operations, drastically reducing memory footprint and deployment costs.
• ☁️ Cloud Vector Storage: Uses Supabase pgvector for lighting-fast semantic search across ~114,000+ medical data chunks.
• 💬 Dual-Response System: Provides answers strictly from the verified medical knowledge base alongside a generalized AI response.
• 🔒 Secure User Sessions: Built with a Custom Django Authentication Model. Uses email-based login (no usernames) and securely saves chat histories in a PostgreSQL database.
• ⚙️ Dynamic Admin Control: Switch between LLM Providers (Gemini, OpenRouter, Groq) dynamically via the Django Admin panel using predefined dropdown choices.
• 🚀 Production Ready: Fully configured with gunicorn, whitenoise, and deployment scripts for seamless hosting on platforms like Render or Railway.

---

🛠️ Tech Stack

• Backend Framework: Django 5.2
• LLM Orchestration: LlamaIndex
• Generative AI: Google Gemini (2.5 Flash / 2.5 Pro)
• Embedding Model: Hugging Face Inference API (BAAI/bge-small-en-v1.5)
• Vector Database: Supabase (pgvector via vecs)
• Relational Database: Supabase PostgreSQL (via dj-database-url)

---

📂 Project Structure

MedRec/
├── accounts/               # Django app for user authentication
├── app/                    # Main Django app handling chatbot and RAG logic
│   ├── static/             # CSS, JS, and local FAISS vector indices
│   └── views.py            # Chatbot API endpoints and LLM fallback logic
├── data/                   # Directory for storing raw medicine CSV datasets
├── MedRec/                 # Django project settings and root configurations
├── templates/              # HTML templates (accounts and chatbot UI)
├── .env                    # Environment variables (API keys, DB URLs)
├── build.sh                # Deployment build script for Render
├── manage.py               # Django command-line utility
├── requirements.txt        # Python dependencies
└── upload_to_supabase.py   # One-time script for ingesting data to Supabase

---

⚙️ Prerequisites

Before you begin, ensure you have the following API keys and URLs ready:

1. Google AI API Key (For Gemini LLM)
2. Hugging Face Access Token (Fine-grained token with Inference API permissions)
3. Supabase Database URL (PostgreSQL connection string with a valid password)

---

🚀 Installation & Local Setup

1. Clone the repository

git clone https://github.com/ArkaKarmoker/MedRec.git
cd MedRec

2. Create a Virtual Environment

python -m venv venv
source venv/bin/activate  # On Windows use: .\venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

4. Configure Environment Variables

Create a .env file in the root directory and add your credentials:

GEMINI_API_KEY=your_gemini_api_key_here
HF_TOKEN=your_huggingface_fine_grained_token_here
SUPABASE_DB_URL=postgresql://postgres.xxx:yourpassword@aws-xxx.pooler.supabase.com:5432/postgres
SECRET_KEY=your_django_secret_key_here
DEBUG=True

5. Run Database Migrations

Since Django is connected to Supabase PostgreSQL, this will create the necessary auth and session tables in the cloud.

python manage.py migrate

6. Start the Server

python manage.py runserver

Visit http://127.0.0.1:8000/ in your browser.

---

📊 Dataset Ingestion (One-time Setup)

If you have a new CSV dataset of medicines and need to upload it to Supabase for the very first time:

1. Place your CSV file in the data/ folder.
2. Update the file path in upload_to_supabase.py.
3. Run the ingestion script (This will take some time depending on dataset size):

python upload_to_supabase.py

Note: This script generates embeddings locally via SentenceTransformers and saves them to both Supabase and a local FAISS index (as backup).

---

🌍 Deployment (Render.com)

This project is fully optimized for free-tier deployments on Render.

1. Create a New Web Service and connect this GitHub repository.
2. Set the Build Command to:

   ./build.sh

3. Set the Start Command to:

   gunicorn MedRec.wsgi:application

4. Add all environment variables from your .env file into Render's Environment Variables dashboard (Set DEBUG to False).
5. Click Deploy.

---

© 2026 Arka Karmoker. All rights reserved.