LifeSync

An iPhone app for logging meals by photo, journaling and tracking the rest of your day, with AI that estimates nutrition and summarizes how the day went.

The app started as "JournalX" in February 2026 and was renamed LifeSync before the TestFlight beta. Some internal names (the Expo slug JournalX2, the @journalx_ storage prefix) still use the old name so existing installs keep their data.

Beta: join on TestFlight

What it does

• Food logging by photo. Snap a meal, pick a photo or type what you ate. The app returns a name, portion, a per-item breakdown, calories, protein, carbs, fat, fiber, sugar, sodium, a 1 to 10 health score and a one-line tip. If a nutrition label is in the photo it copies the label. For branded or restaurant food it uses the published numbers. Otherwise it estimates portions from visual cues.
• Snap and go. Logging never waits on the AI. Each meal goes into a local queue and is estimated in the background (up to 3 at a time). Without a connection, meals wait and retry later, and they keep the time you actually ate.
• Corrections. Add a note like "ate half" to a logged meal and it gets re-estimated from the earlier analysis.
• Daily targets and coaching. Targets come from the Mifflin-St Jeor equation, your activity level and a Gain / Maintain / Lose goal. The Today card shows progress, a day grade and an AI suggestion for the next meal or snack that closes the biggest gap.
• The rest of the day. Journal entries, mood, water, sleep, exercise (with AI fill from a text description), meds and other events, on a calendar day view with draggable blocks. Swipe between days on the home screen.
• AI summaries and insights. A short daily summary on the home screen and an Insights tab that looks across recent days.
• Reminders. Push notifications for a missing meal log, journal entry or low water.
• Accounts. Email and password, Sign in with Apple, or Google. Password reset with a 6-digit code in the app. Account deletion from Settings removes the user and all their rows.

How it works

iPhone app (Expo / React Native)
   |  Supabase session token
   |-------------------------------> Supabase (Postgres + Auth, row level security)
   |
   |  Authorization: Bearer <token>
   '-------------------------------> Vercel serverless API (api/) ---> OpenRouter (AI models)

App (App.js, src/). Expo SDK 54, React Native 0.81, React Navigation tabs (Home, Insights, Add, Calendar, Settings). One store (src/store/useStore.js) holds the user's data, caches it on the device and syncs it to Supabase (src/api/supabaseSync.js). Photos are resized to 1024 px wide JPEGs before upload. The client only holds the Supabase URL and anon key, which are public by design.

API (api/, deployed on Vercel):

Endpoint	What it does	Daily cap
POST /api/ai/food-photo	Photo, typed meal or correction to structured nutrition	40 per user
POST /api/ai/food-coach	One next-meal idea from today's totals and targets	30 per user
POST /api/ai/extract	Exercise details from a sentence
POST /api/ai/summary, /api/ai/summary-short	Day summary
POST /api/ai/insights	Insights over recent events, journal and moods
POST /api/notifications/register	Saves the Expo push token for the signed-in user
GET /api/notifications/cron	Daily reminder job (Vercel Cron, checks CRON_SECRET)
GET /api/udid	Device-ID sign-up page used for the ad hoc beta before TestFlight

Every AI endpoint goes through api/_lib/auth.ts, which verifies the Supabase session token, applies a 30 requests per minute per user limit and counts daily caps in the user's Supabase app_metadata. AI keys only exist as Vercel environment variables.

AI providers. All model calls go through OpenRouter. Food uses google/gemini-3.8-flash with openai/gpt-6-sol as an automatic fallback, with strict JSON schema output (api/_lib/openrouter.ts). Summaries, insights and exercise fill use openai/gpt-4.1-mini. Food photo logging first ran on Claude and moved to OpenRouter in October 2026.

Database (supabase/). Tables for events, journal entries, summaries, mood, profile, water, push tokens and a notification log. Every table has row level security with owner-only policies, so a user's token can only read and write their own rows. Schema changes are numbered SQL migrations (001 to 008). The delete-account edge function deletes the auth user and foreign key cascades remove their data.

More detail and the main design decisions: HOW-IT-WORKS.md.

Engineering notes

• Tests. npm run test:nutrition (31 assertions: goal math, height parsing including iOS smart quotes, food fields, day totals and the day grade) and npm run test:store-merge (4 assertions: merging server and local events, aborting a refresh when one table fails to load). Both run in plain Node with node:assert, because the logic lives in small CommonJS modules (src/utils/nutrition.js, src/utils/goalSuggest.js, src/store/eventMerge.js).
• Backward compatible API. Friends ran older builds while the API changed, so new endpoints return a superset of the old response and keep the old paths. Migration 008 (an events.nutrition jsonb column and profile goal fields) had to be applied before a new build could sync.
• iOS 27 launch fix. Apps built with the iOS 27 SDK must use the UIScene lifecycle, and Expo 54's template does not. plugins/withSceneDelegate.js is a config plugin that adds a SceneDelegate during expo prebuild.
• Size. About 17,000 lines of JavaScript, TypeScript and SQL, 46 commits from February to October 2026.

This is the beginning of the README. Read the rest at the original. Read the full README ↗