shifa.ai - AI Health Assistant

A production-ready MVP for an AI-powered health assistant using FastAPI, SQLAlchemy (SQLite), and Google Gemini Pro.

🚀 Preview

A clean, modern interface designed for clarity, trust, and usability.

Features

• Modern UI: Clean, responsive design with dark mode support.
• User Onboarding: Simple flow to collect basic user info.
• AI Chat: Powered by Google Gemini with health-specific guardrails.
• 🎙️ AI Voice Assistant: Real-time, voice-to-voice health receptionist — answers questions about clinic services / hours / pricing from a knowledge base (RAG) and books appointments live, then ends the call automatically. Open at /voice.
• Chat History: Messages are saved and reloaded automatically.
• Usage Limits: Configurable limit on free questions (default: 10).

Tech Stack

• Backend: Python 3.10+, FastAPI, SQLAlchemy, SQLite, Google Generative AI.
• Voice: Google Gemini Live (native speech-to-speech) · Pipecat (real-time audio pipeline) · LangGraph (stateful agent) · RAG with Gemini embeddings · WebSockets.
• Frontend: Tailwind CSS (CDN), Material Symbols, Vanilla JavaScript, Web Audio API.

Setup

1. Clone the project

2. Create a virtual environment: Windows:

   python -m venv venv
   venv\Scripts\activate

   macOS/Linux:

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Configure Environment Variables: Update the .env file with your Gemini API Key:

   GEMINI_API_KEY=your_actual_api_key_here

Running the Project

Start the server using uvicorn:

uvicorn backend.main:app --reload

The application will be available at http://127.0.0.1:8000.

• Chat assistant → /get-started
• 🎙️ Voice assistant → http://127.0.0.1:8000/voice

🎧 Use headphones for the voice assistant. On open speakers the mic picks up the assistant's own voice (echo) and the conversation degrades. Headphones fix this.

---

Folder Structure

• backend/: FastAPI application code.
  • backend/voice/: Real-time voice assistant (Gemini Live + Pipecat + LangGraph + RAG).
• frontend/: Static HTML/JS files (voice.html = the voice UI).
• shifa.db: SQLite database file (generated on first run).

👨‍💻 Meet the Developer

Hi, I'm Hassan Khan 👋
AI Engineer | RAG Systems | LLM Chatbots | AI Agents

I build production-grade AI systems that help businesses:

• RAG Systems, Custom LLM Chatbots, AI Agents, Python based AI Backends
• Chat with their private data
• Automate workflows
• Deploy scalable, reliable AI solutions

---

🤝 Let’s Connect

📧 Email: hassanaiengineer@gmail.com
🔗 LinkedIn: https://www.linkedin.com/in/hassan-khan-4961b722b/
🔗 Upwork: https://www.upwork.com/freelancers/~016ca6a619d9683838