🏥 HealthDesk Pro

Privacy-First Offline Medical AI with Advanced Reasoning

Features • Architecture • Installation • Usage • Team

---

🚨 Problem

People search symptoms on Google and get:

• Wrong medical advice
• Ads instead of answers
• Privacy leakage to cloud servers

No offline, private, reliable medical AI exists for rural + low-internet areas.

---

💡 Solution

HealthDesk Pro is a fully offline medical AI system that:

• Runs locally (no internet required)
• Uses WHO / NIH documents only
• Gives cited medical answers
• Detects emergencies instantly
• Supports multiple Indian languages

---

✨ Features

🔒 Privacy First

• 100% offline processing
• No cloud APIs
• No data tracking

🧠 Medical Intelligence

• RAG-based document QA
• WHO / NIH citations
• Confidence scoring

🚨 Emergency Detection

• Chest pain → instant emergency alert
• Stroke detection
• Breathing difficulty alerts
• Auto suggests calling 108 (India)

🌐 Multilingual

• English
• Tamil
• Hindi
• Telugu
• +200 languages via models

---

🏗️ Architecture

Frontend (React) ↓ FastAPI Backend ↓ RAG Engine (ChromaDB) ↓ Ollama Models ↓ WHO / NIH Documents

---

🛠️ Tech Stack

• Frontend: React 18
• Backend: FastAPI
• LLM: Ollama
• Models: Gemma 3 / DeepSeek / Qwen
• Vector DB: ChromaDB
• Embeddings: nomic-embed-text

---

📦 Installation

1. Clone Repo

git clone https://github.com/mithilP007/Health-desk.git
cd Health-desk
2. Install Ollama Models
ollama pull gemma3:4b
ollama pull deepseek-r1:14b
ollama pull qwen2.5:14b
ollama pull nomic-embed-text
3. Backend Setup
cd backend
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
uvicorn main:app --reload
4. Frontend Setup
cd frontend
npm install
npm start
🚀 Usage

Open:

http://localhost:3000

Example queries:

"What are dengue symptoms?"
"I have chest pain"
"How does malaria spread?"
"நெஞ்சு வலி இருக்கிறது"
📁 Project Structure
Health-desk/
├── backend/
├── frontend/
├── knowledge/
├── docker-compose.yml
├── README.md
🧪 Testing
pytest backend/tests/

Check:

Emergency detection
Citation accuracy
Safety filters
Multilingual responses
👥 Team
Role	Work
Engine Dev	Backend + RAG + API
Knowledge Engineer	WHO/NIH documents
Safety Engineer	Prompts + testing
🔮 Future Scope
Voice AI (speech-to-text)
Mobile app
Medical image detection
IoT health devices integration
⚠️ Disclaimer

This system provides general medical information only.
Not a replacement for professional doctors.

For emergencies → Call 108 (India) immediately.

❤️ Credits
WHO (Medical Guidelines)
NIH (Health Data)
Ollama (Local LLMs)
Open-source community
<p align="center"> Built for rural healthcare accessibility ❤️ </p> ```