How tilakjain619/medtalk works
Project type and purpose
• MedTalk is a full-stack web application: an AI-powered healthcare assistant.
• Purpose: Users upload prescription images; the app extracts medication details via OCR, generates a patient-friendly summary, and provides a conversational chat powered by an LLM to answer follow-up questions.
File structure
• Top-level folders:
• client/ – React-based front end
– public, src, components, lib, config files (tsconfig, eslint, Vite).
• server/ – Node.js + Express back end
– app.js entry point, routes (ai.js), utils (OCR & OpenRouter integration), no persistent database.
• Key patterns:
• SPA (single page app) in client
• REST-style API endpoints in server/routes
• External service wrappers in server/utils
• Deployment/config files:
• client/vite.config.mjs, server/package.json, environment variable placeholders (.env for VITE_BACKEND_URL, OPENROUTER_API_KEY).
README for additional insights
• Frontend uses React + Tailwind CSS.
• Backend uses Express.
• OCR via Puter.ai.
• AI chat via Google Gemma accessed through the OpenRouter API.
• No database indicated—stateless, real-time processing pipeline.