An iOS calorie tracker. Search 988,000 Open Food Facts products offline — no account, no server, no network.

I built it because MyFitnessPal had become unbearable. Not the features — the speed. Opening the app took several seconds, logging a meal took several more. I wanted something that opens instantly and lets me log breakfast before the coffee is done.

This repo is entirely vibe-coded. I did not write a single line of it. All of it — the 4,900 lines of Swift, the Rust tools, the import scripts, the 87 commits — came out of Claude Code. My part was describing what I wanted, running it on my phone, and saying what felt wrong.

The UI is in French only, and so is the product database.

Dashboard	Search	Product detail

Weight	Meal detail	Goals

Food log

• Four meals a day: breakfast, lunch, snack, dinner
• Remaining-calorie ring, macro bars (protein, carbs, fat, fiber)
• Live meal and day totals while you pick a portion size
• Move between days with arrows or a calendar
• Edit or delete any logged food
• Multi-select to delete several entries at once or copy them to another meal
• One-tap copy of the previous day's meal into an empty section
• Per-meal detail screen with aggregated nutrients

Food search

• Full-text search over the French Open Food Facts database, bundled in the app (SQLite + FTS5)
• Ranked by relevance then popularity, 150 ms debounce, run off the main thread
• EAN barcode scanning through the camera
• History of recent and most-used foods, filterable
• Default portion pre-filled from the product's packaged quantity
• Manual entry: type grams and macros directly, no product needed
• Custom foods, created and stored locally
• Fix incomplete Open Food Facts entries: your values override the bundled database

Weight

• Log today's weight straight from the dashboard
• Trend chart over 7, 30, 90 days or all time, Catmull-Rom smoothed
• Weight goals (loss, gain, maintain) with the delta shown on the dashboard

Goals and data

• Daily targets for calories, protein, carbs, fat, sugar, salt
• Fiber target derived automatically (15 g per 1,000 kcal)
• Manual export and import of the database, with a reminder when the last export is over 7 days old

• SwiftUI, Swift 6 with strict concurrency, iOS 17+
• GRDB over two SQLite databases: Open Food Facts read-only, user data read-write
• FTS5 for full-text search, unicode61 tokenizer with diacritics stripped
• Swift Charts for the weight curve, AVFoundation for barcode scanning
• XcodeGen — the Xcode project is generated from project.yml
• Two Rust tools to build the product database
• No network calls, no account, no telemetry. Everything stays on the device.

The visual theme borrows from Duolingo: saturated colors, rounded corners, heavy rounded type, spring animations.

The app bundles a 988,000-product SQLite file. It is not versioned — it weighs 230 MB. Build it in two steps.

1. Download the Open Food Facts JSONL dump:

curl -o data/openfoodfacts-products.jsonl.gz \
  https://static.openfoodfacts.org/data/openfoodfacts-products.jsonl.gz
gunzip data/openfoodfacts-products.jsonl.gz

2. Reduce it to French products, then convert it to SQLite:

cargo run --release --manifest-path tools/off-reduce/Cargo.toml -- \
  data/openfoodfacts-products.jsonl data/off_fr.jsonl

cargo run --release --manifest-path tools/off-to-sqlite/Cargo.toml -- \
  data/off_fr.jsonl data/off_fr.sqlite

off-reduce keeps products sold in France and drops every field the app does not read. off-to-sqlite builds the table and its FTS5 index.

The build also expects data/user.sqlite, the seed user database. An empty file works: the app copies it into Application Support on first launch and never touches it again.

brew install xcodegen
xcodegen generate

xcodebuild -scheme kcalz -project kcalz.xcodeproj \
  -destination 'platform=iOS Simulator,name=iPhone 16 Pro' build

To install on a device, swap the destination for your device id (xcrun xctrace list devices), add -allowProvisioningUpdates, then:

xcrun devicectl device install app --device <DEVICE_ID> <path>/kcalz.app

Set your own DEVELOPMENT_TEAM in project.yml before signing.

kcalz/
  App/          entry point, store bootstrap
  Models/       data types
  Stores/       OFFStore (read-only), UserStore (read-write)
  ViewModels/   @Observable
  Views/        one view per screen, shared components
  Utils/        theme, date formatting
tools/
  off-reduce/     Rust — OFF dump → filtered JSONL
  off-to-sqlite/  Rust — JSONL → SQLite + FTS5
  mfp-import/     TypeScript — import a MyFitnessPal export
docs/
  SPECS.md      original spec
  plans/        one design document per feature

docs/plans/ tracks the design feature by feature. It is the most readable trace of how the app was built.

• French only, both the UI and the product data
• No iCloud sync — backups go through the manual export
• Multi-select opens from a context menu rather than a direct long press (SwiftUI gesture conflict with the scroll view)
• No automated tests

MIT. Product data comes from Open Food Facts, under the ODbL license.