VITARA — AI Nutrition & Fitness Companion
- Nuxt 4
- Vue.js
- TypeScript
- PostgreSQL
- Drizzle ORM
- Gemini API
- PWA
Full-stack Nuxt 4 app built around Indonesian eating habits: AI food scanning and an AI coach powered by Google Gemini, a budget-aware meal planner, and RPG-style workout progress — installable as a PWA.
Private repository — live demo available
VITARA is a full-stack web app that helps people build healthier habits around Indonesian food. Users log meals from a local food catalog in grams, millilitres, or servings, track water intake and weight trends, and get daily energy targets from an eight-step onboarding based on the Mifflin-St Jeor formula with built-in safety limits.
Google Gemini powers two features: food scanning from a photo or a text description — every candidate must be confirmed by the user, with per-feature consent, quotas, and a manual fallback — and an AI coach with chat and voice modes that screens for red flags before the model is called. A deterministic optimizer builds budget-friendly meal plans and shopping lists, and workouts feed an RPG layer of XP, levels, streaks, quests, and a 7-day muscle map.
Built with Nuxt 4, Nuxt UI, and TypeScript on a clean architecture: use cases and ports in the core, with adapters for Drizzle/PostgreSQL (row-level security), Gemini, storage, mail, and push. It ships as an installable PWA with web-push reminders, plus a Capacitor build and a REST API with OpenAPI docs for native clients. The repository is private; the live app is open at vit-fit.vercel.app.