Welva, an iOS nutrition and workout app in SwiftUI

Welva, an iOS health and strength training app, started as FitTrack in February 2025, out of a personal need: tracking weight, nutrition and workouts without juggling several apps that don’t talk to each other. The first version, built with React Native and Expo, already read Apple Health and calculated nutrition goals.

The project grew into a product published on the App Store. In July 2026 FitTrack became Welva, and the app was rewritten in native Swift for its V4, with an Apple Watch app, a Welva Plus subscription and a marketing site at welva.app.

A health app is only as good as its numbers. Calories had to come from a reliable source, a workout recorded on the Watch had to appear only once in Health, and sync had to survive a reinstall or a change of device.

AI raised a different problem: the coach, meal photo analysis and reports had to rely on computed data, not make it up. The app also had to pass Apple's review, which requires accurate privacy declarations and a flawless purchase flow.

I rewrote the app in SwiftUI, split into packages, with a Watch app that counts reps and learns machine movements from confirmed sets. The catalogue holds 425 illustrated exercises, a muscle map that lights up based on the session, voice entry for sets ("a hundred kilos, ten reps"), widgets and Live Activities on the lock screen.

On the server side, a Cloudflare Worker (Hono, D1, KV, R2) handles Sign in with Apple, category-based sync, GDPR export and an MCP server for connecting your data to an AI assistant. Food search combines ANSES's Ciqual 2025 table with Open Food Facts using Workers AI embeddings in Vectorize, and meal photo analysis runs in the background on Cloudflare Workflows. Subscriptions go through RevenueCat, and the API contract is generated as OpenAPI from the zod schemas.