Metra Health AI
GLP-1 Medication & Wellness Tracker
About the Project
Metra Health AI is a live, cross-platform mobile app for people on GLP-1 medications such as Ozempic, Wegovy, Mounjaro and Zepbound. The app supports 19 GLP-1 variants — injectable and oral — alongside daily logging for weight, side effects, food and macros, water, exercise and reminders. Two AI surfaces sit on top of that data: a conversational health chat with a retrieval-augmented context window over the user's recent logs, and a daily AI insight feed that generates 3–5 personalized cards. A dedicated food backend backs nutrition logging with the USDA FoodData Central database, full-text and vector search, and barcode and photo entry. The app is on the App Store and Google Play with paying subscribers managed through Superwall.
The Challenge
GLP-1 patients juggle a dosing schedule, side-effect patterns, weight changes, and a strict nutrition signal — across multiple apps, paper notes, and provider portals. Metra needed to bring all of those signals into one product without becoming a generic health tracker, and layer AI on top in a way that actually understood the patient's recent history rather than answering in generalities. Doing that meant ingesting and normalizing data from Apple Health and Samsung Health Connect, building a food database that doesn't lose users at the search bar, supporting 19 distinct medications with different cadences and dose ladders, and shipping AI features behind a token-aware free tier without breaking the paywall economics.
Our Approach
We built the app in Flutter so iOS and Android share one codebase, with flutter_bloc and hydrated_bloc for offline-friendly state. The backend runs on Supabase: TypeScript Edge Functions for chat, insights, ingestion and ops, plus a Postgres schema managed strictly through migrations, with pgvector powering RAG. AI Chat and AI Insights both run on OpenAI GPT-4.1-mini, with the chat function pulling the user's most recent logs and a curated knowledge base into context, and Insights running once per day to generate 3–5 prioritized cards. Food logging is served by a separate Go and PocketBase service over a USDA / FDC food database with full-text and Qdrant vector search, supplemented by FatSecret and a photo-food OpenAI-vision pipeline. Subscriptions and entitlements run through Superwall, synced into Supabase so token limits and feature gating stay consistent across stores.
The Product in Action
Tech Stack
The technologies and tools that powered this project.
Project Timeline
Foundations
Phase 1Flutter app skeleton, Supabase schema and auth, dose, weight and side-effect logging, and the first cut of the food search backend on Go and PocketBase backed by USDA FDC data.
AI Chat & Insights
Phase 2Edge function with RAG over the user's recent logs and a curated knowledge base, daily AI Insights generating 3–5 prioritized cards, and the supporting prompt and embedding pipeline.
Health Integrations & Subscriptions
Phase 3Apple Health and Samsung Health Connect ingestion, Superwall paywall and entitlement sync, token-aware free tier, photo and barcode food entry, and freemium alerting.
Live Operations
OngoingApp Store and Google Play release, ongoing iteration on the chat and insights prompts, longitudinal data window expansions, and continuous schema migrations.
Team Composition
Founder/CEO, Co-Founder, 1 Backend Engineer, 1 Mobile Engineer, 1 Designer, plus Above Software engineering and infra advisory
Let's build something great together
Ready to turn your idea into a product users love? Book a discovery call and let's get started.
Book a Call