Live transcription tool that predicts likely diagnoses and prompts the next 3 most relevant follow up questions to narrow further.

A React-based medical assistant application that uses OpenAI’s realtime API for voice interactions and provides clinical note extraction, diagnostic suggestions, and structured medical documentation.

Project Structure Frontend: React SPA with Vite build system Backend: Express.js server with OpenAI realtime API integration Deployment: AWS SST framework Database: DynamoDB (via SST) Authentication: JWT-based with bcrypt password hashing Key Components Client (/client/) App.jsx - Main application component AuthWrapper.jsx - Authentication wrapper DiagnosisPanel.jsx - Medical diagnosis interface TranscriptionPanel.jsx - Real-time transcription display StructuredNotePanel.jsx - Clinical note formatting SuggestedQuestionsPanel.jsx - AI-generated follow-up questions SessionControls.jsx - Audio session management Server (/server.js) Express server with Vite middleware OpenAI realtime API integration JWT authentication endpoints Clinical note processing APIs Services (/services/) clinicalNoteExtractor.js - Extracts structured clinical data diagnosticEngine.js - Provides diagnostic suggestions Infrastructure (/stacks/) API.ts - API Gateway configuration Database.ts - DynamoDB tables Web.ts - Static site hosting Available Commands Development npm run dev - Start development server npm run sst:dev - Start SST development environment Build & Deploy npm run build - Build client and server npm run build:client - Build React client only npm run build:server - Build server for production npm run sst:deploy - Deploy to AWS npm run sst:remove - Remove AWS resources Code Quality npm run lint - Run ESLint with auto-fix Environment Variables Required environment variables:

OPENAI_API_KEY - OpenAI API key for realtime sessions JWT_SECRET - Secret for JWT token signing SESSION_SECRET - Express session secret Authentication The app uses JWT tokens for authentication with the following endpoints:

POST /login - User login POST /register - User registration Protected routes require Authorization: Bearer header OpenAI Integration Uses OpenAI’s realtime API model: gpt-4o-realtime-preview Voice mode: “verse” Generates ephemeral tokens for client-side connections Supports real-time audio transcription and responses Medical Features Clinical Note Extraction: Processes conversation transcripts into structured medical notes Diagnostic Engine: Provides AI-powered diagnostic suggestions Structured Documentation: Formats clinical encounters into standardized formats Suggested Questions: Generates relevant follow-up questions for medical interviews UI Framework shadcn/ui Integration (Updated 2025-01-07) Migrated from custom components to shadcn/ui component library Consistent design system with modern card-based layouts Professional medical application styling Key UI Components AppLayout.jsx - Main layout wrapper with sidebar integration AppSidebar.jsx - Collapsible sidebar navigation with medical-focused sections All panels now use shadcn/ui Cards, ScrollAreas, and Badges Responsive design with mobile support via use-mobile hook Layout Structure Sidebar Navigation: Medical console navigation with user profile and logout Main Content: Full-width responsive grid layout No Header: Clean design with all controls in sidebar User Management: Logout functionality integrated in sidebar footer shadcn/ui Components Used Card, CardContent, CardHeader, CardTitle Button with variants Badge ScrollArea Avatar Separator Sidebar components Development Notes Uses ES modules (“type”: “module” in package.json) React 18 with modern hooks shadcn/ui + Tailwind CSS for consistent design system Vite for fast development and building TypeScript support for infrastructure code