Archimed - Medical Transcription & Note Generation Platform

A real-time medical transcription platform that converts doctor-patient conversations into structured medical notes using AI.

Architecture Overview

Frontend (Nuxt.js/Vue.js)

• Framework: Nuxt 3 with Vue.js components
• Features: Real-time audio recording, WebSocket communication, payment integration (Stripe)
• Authentication: Firebase Auth with session cookies
• Internationalization: English/French support
• Key Components: AudioRecorder, SessionList, MedicalNoteDisplay

Payment System (Stripe Integration)

• Subscription Management: Trial activation with payment collection
• Checkout Sessions: Server-side checkout session creation with webhook validation
• Payment History: Customer payment tracking and subscription status monitoring
• API Endpoints: /api/payments/create-checkout-session, /api/subscription/status
• Webhook Handling: Secure Stripe webhook processing for payment events
• Components: PaymentModal, SubscriptionInfo, TrialPaymentCollection

WebSocket Service

• Real-time bidirectional communication between client and transcription server
• Connection management with automatic reconnection and heartbeat monitoring
• Message queuing for offline scenarios
• Client identity management with persistent session tracking

Web Worker (Audio Processing)

• File: client/public/workers/audioProcessor.worker.js
• Handles audio chunk processing and IndexedDB storage
• Manages offline audio buffering and synchronization
• Implements automatic cleanup and storage optimization

Python Backend (FastAPI)

• Location: transcription/ directory
• WebSocket Server: Real-time audio processing at localhost:8080
• Features: Session management, audio transcription, medical note generation
• AI Models: Faster-Whisper for transcription, custom medical note generation

Whisper Integration

• Model: Faster-Whisper "turbo" model with CUDA acceleration
• Real-time streaming transcription with beam search (beam_size=5)
• Optimized for medical terminology and French/English languages

Data Flow

1. Browser captures audio → Web Worker processes chunks
2. WebSocket sends audio to Python backend
3. Faster-Whisper transcribes audio in real-time
4. AI generates structured medical notes
5. Results streamed back to client via WebSocket

Setup

Install dependencies for both client and transcription services:

# Client setup
cd client && npm install

# Python backend setup
cd transcription && pip install -r requirements.txt

Development

Start both services:

# Start transcription server (port 8080)
cd transcription && python run.py

# Start Nuxt development server (port 3000)
cd client && npm run dev