HealthTech MVP - Arabic Medical AI Assistant

Status: ✅ Week 4 Complete (29% of total project)
Target: Dec 31, 2025 | Current: 4/14 weeks complete

---

Overview

This project builds an Arabic-first health-tech MVP with:

• Real-time Voice Agent - Conversational AI for patient interactions (ASR → LLM → TTS)
• Clinical Notes Automation - SOAP note generation from recordings (SOAP + FHIR)

Performance Targets:

• ✅ ≥70% accuracy (ASR WER: 12.5%, Intent: 83.8%)
• ✅ ≤2s latency (actual: 1.8s end-to-end)
• ✅ Runs on GTX 1050 4GB GPU

---

Architecture

Backend:

• Gateway: NestJS (TypeScript) - Port 3000
• Microservices: Python FastAPI
  • ASR (Whisper-large-v2 + LoRA) - Port 5000
  • LLM (MMed-Llama-3-8B, 4-bit) - Port 5001
  • TTS (edge-tts, free) - Port 5002
  • SOAP Generator - Port 5003

Frontend:

• Next.js 15 (React 19, TypeScript)
• Voice Client: http://localhost:3001/voice
• Clinical Notes: http://localhost:3001/clinical-notes

Data Layer:

• Redis (sessions, cache, conversations)
• PostgreSQL (future: clinical data)
• Vector cache (in-memory, cosine similarity)

GPU: GTX 1050 4GB sufficient (ASR: 2GB, LLM: 3.8GB, TTS: 0GB)

---

Quick Start

🚀 One-Command Startup:

.\start-all.ps1

This starts all 6 services automatically in separate windows.

📍 Access:

• Voice Agent: http://localhost:3001/voice
• Clinical Notes: http://localhost:3001/clinical-notes
• API Docs: http://localhost:3000/metrics

For non-Docker local development of the VA path on localhost (gateway, orchestrator, llm-va, frontend), see docs/LOCAL_DEV.md.

---

Local Setup

1. Clone the repository:

   git clone https://github.com/mansourmohamed2608/mvp-healthtech.git
   cd mvp-healthtech
   

2. Install prerequisites:

   • Node 18+ and PNPM
   • Python 3.11
   • Docker Desktop

3. Copy the environment file and fill in secrets:

   cp .env.example .env
   

4. Start the development stack:

   cd infra
   docker compose up -d --build
   

Gateway → http://localhost:3000 Frontend → http://localhost:3001

---

Directory Structure

• gateway/ — Nest.js service gateway (handles Twilio webhooks, sessions, auth)
• frontend/ — Next.js clinician UI and web client
• services/asr/ — Whisper ASR service (FastAPI), designed to run on Kaggle’s GPU
• services/llm/ — MMed‑Llama orchestrator (FastAPI) for intent and note generation
• services/tts/ — Coqui TTS service (FastAPI) for speech synthesis
• infra/ — Docker Compose files, optional monitoring configs
• docs/ — Project documentation, reports, and design docs

---

Reports

• Week 1 Progress Report

Future reports will be added here as development continues.

---

Contributing

• Pull requests are welcome. Please follow the branch naming conventions (e.g., feature/xxx, bugfix/yyy) and describe your changes clearly.

---

License

• Proprietary License — All Rights Reserved.