Product Engineering · Companion AppPrivate / demonstration companion

MY HEALTHY LIVING — LIVING GUIDE

My Healthy Living

Personal living guide for planning, cooking, shopping and progress

A daily companion for planning, cooking, shopping, and progress, with a handbook as the content source of truth. Meal planning is a deterministic engine, not a generative model. The live app is provided for private and demonstration use. It is not clinical software and does not offer diagnosis, treatment, or medical advice.

Next.js 15TypeScriptTailwindSupabaseVercel

The Challenge

The engineering problem

A household needed one place for daily meals, a week plan, shopping, recipes, and light progress tracking — without turning meal planning into an opaque model, and without moving the handbook PDF pipeline onto the web runtime.

Solution

What EWS built

The companion is a Next.js App Router PWA. Handbook YAML remains the content source; a sync step emits JSON for the app. Domain logic for meals, shopping, and “what can I cook” is deterministic. Supabase provides magic-link authentication and Postgres when configured. Resend plus a Vercel hourly cron deliver notification routines. The A4 handbook PDF pipeline stays local and is not the Vercel architecture.

Product experience

Working software

Selected application views. These are product screens, not conceptual placeholders.

Today — cooking, planning, shopping, and progress

Architecture

How the system is composed

The web companion and the local handbook pipeline share content. They do not share a runtime.

Companion application

User

Today / Cook / Plan / Shop / Progress

Next.js PWA

App Router, TypeScript, Tailwind

Deterministic planning engine

meal-engine · shopping · cook-match

Supabase Auth

Magic link

PostgreSQL

Household and progress data

Notifications

Resend + Vercel Cron

Capabilities

What was implemented

Today

Daily focus, meals, and shortcuts into cook, plan, shop, and progress.

Cook, plan, shop

Recipe cooking, weekly planning, and aggregated shopping lists from the plan.

Discover and handbook

Recipe and ingredient browsing, plus handbook chapters synced from the same content source.

Progress and settings

Progress tracking and notification preferences. Pattern copy is non-medical.

Magic-link authentication

Sign-in sends a secure link. Authenticated routes are gated in the deployed app.

Engineering decisions

Choices that matter

Deterministic planning, not an LLM

Week generation, replacements, and cook-matching are rule engines. That keeps dietary defaults inspectable.

Handbook remains the source of truth

Recipes and plans live in the handbook data tree. The web app syncs; it is not a second hand-maintained database.

PDF pipeline stays off Vercel

PRESS/PRINT generation uses local Python tooling. The hosted app must not run that pipeline.

Technology

Stack by layer

Frontend
Next.js 15 App RouterTypeScriptTailwindPWA
Application
Deterministic meal engineShopping listsCook matching
Data / Auth
Supabase AuthPostgreSQL
Infrastructure
VercelVercel CronResend
Testing
Vitest

Evidence

What we can stand behind

Implemented product surfaces

Today, Cook, weekly planning, shopping, Discover/recipes, Progress, ingredients, handbook, settings, magic-link auth, and notification routines are implemented in the companion.

Hosted demonstration

The app is deployed for private and demonstration use. That is a product status, not a clinical or consumer-health claim.

What is not claimed

No diagnosis, treatment, medical advice, or weight-loss outcomes. Household identity from internal screenshots is not published.

Related capabilities

Related EWS Capabilities

Discuss a similar project

Start from the workflow, the constraints, and the systems that have to stay authoritative.

Discuss a Similar Project