Yet another workout logger.

A single-purpose iPhone app for logging barbell training: what you lifted, for how many reps, at what RPE — and what to do about it next session. No account, no sync, no subscription. Everything lives on the device.

It exists because a workout log has to work in a gym basement with no signal, with sweaty hands, in the ninety seconds between sets. That rules out most web apps, and the rest-timer requirement rules out all of them: the alert has to fire with the phone locked in a bag.

Logging. Pick a training day from your template, and it pre-fills each exercise with the set/rep structure you used last time. Enter weight, reps, RPE. The set grid is the whole interface.

Tells you what to lift. Every strength exercise carries a suggestion built from your own history. Give an exercise a rep range and an RPE ceiling and it runs double progression — fill the range at or under the cap, then add weight. Without a target it falls back to an estimated-1RM trend heuristic. Weights round to something you can actually load.

Rest timer that survives a locked screen. Local notification, fires with the app backgrounded. It announces the next set by name and weight, and carries a text-input action — reply with an RPE from the lock screen or an Apple Watch and it logs against the correct set, then starts the next rest period.

Heart rate from HealthKit. Per-set HR is stamped at RPE entry and backfilled at session end from the full sample list, which works around Watch→iPhone sync latency. Fails silently by design — a missing Watch never interrupts a workout.

Progression machinery. Deload mode rewrites the day to 60% of top-set weight capped at two sets, and is excluded from suggestions and PR detection. Linked exercises derive their weight from a fraction of another lift (RDL at 0.75 × deadlift) and re-derive every session. Estimated 1RM is Epley, and drives records, PR detection, and the trend heuristic.

Records, history, export. Best set per exercise by e1RM with a per-exercise trend, history as a list or a calendar, and JSON export/import — your data leaves whenever you want it to.

kg or lb. Weights are stored canonically in kilograms and converted for display, so switching units never rewrites what you logged and your PRs stay comparable. The unit drives stepper increments, plate rounding, and the warm-up ramp's bar weight (20 kg / 45 lb).

Building this needs a Mac. There is no Android target and no hosted version.

All commands run from native-ios/. The repository root holds a stale earlier copy of the web app (index.html, storage.js, notifications.js, package.json) with no src/ directory — npm run dev at the root will fail. Treat the root copies as dead.

git clone [email protected]:jonuts/yawl.git
cd yawl/native-ios
npm install

The ios/ directory is committed, so there is no npx cap add ios step.

npm run dev

Vite on localhost:5173. Best for anything UI — layout, state, the training logic. Capacitor Preferences falls back to localStorage, so data persists between reloads.

What does not work in a browser: local notifications and HealthKit. Both are native bridges. healthkit.js returns null rather than throwing, so the app runs fine without them — you just won't see heart rate or rest alerts.

Note the app is portrait-locked and iPhone-shaped; use a device viewport around 375×812 or the layout will look stretched.

npm run ios      # build + sync + open Xcode

Then in Xcode: select the App target → Signing & Capabilities → tick Automatically manage signing and pick your Apple ID team. Plug in the iPhone, unlock it, select it in the device dropdown, hit ▶.

First run on the phone also needs Settings → General → VPN & Device Management → your Apple ID → Trust.

Other scripts:

npm run build    # vite build -> dist/ (what Capacitor copies in)
npm run sync     # build + cap sync ios

Native changes need Xcode. npm run sync only refreshes web assets. Touch any Swift file, Info.plist, or the entitlements and you must rebuild in Xcode.

On a free Apple ID the app signature expires after 7 days — the app stops opening and you re-run it from Xcode (~1 minute). The $99/year Developer Program raises that to a year and allows TestFlight installs over the air. Worth running free for a couple of weeks before deciding.

native-ios/
├── src/
│   ├── WorkoutTracker.jsx   ~3,000 lines — this IS the app
│   ├── storage.js           Capacitor Preferences behind an async get/set
│   ├── notifications.js     rest-timer notifications + the RPE reply action
│   ├── healthkit.js         heart rate; fails silently on purpose
│   └── index.css            load-bearing iOS scroll fixes — read the comments
├── ios/App/App/
│   ├── HealthKitPlugin.swift   hand-written plugin, not a package
│   └── MyViewController.swift  Capacitor 6 needs plugins registered here
└── branding/                SVG sources for the icon and splash

WorkoutTracker.jsx is one file holding every view, switched on a view string. There is no router and no state library — all state lives in the default-exported component at the bottom, and every view is a pure prop-driven child above it. Adding a feature usually means touching both ends of the file.

CLAUDE.md documents the architecture in more depth: the data model, the persistence keys, and the domain rules worth knowing before editing.

cap sync and CocoaPods. If LANG isn't a UTF-8 locale, CocoaPods dies with Encoding::CompatibilityError. The npm scripts set it; a bare npx cap sync ios from a fresh shell may not.

Don't change the bundle ID. It's com.jonah.workoutlog and the app is named Yawl — that mismatch is deliberate. Changing it makes iOS install a new app and orphan every logged session on the device.

Don't restore scrolling to body. index.css fixes html, body and scrolls #root instead. The WKWebView's own scroll view rubber-bands on iOS and drags position: fixed elements — the rest timer, tab bar, save bars — along with it.

Inputs are pinned to 16px. iOS auto-zooms on focus below that.

There are no tests, linter, or typechecker. Verification means running it. The functions where a silent regression would cost real training data are buildSuggestion, applyDeloadTransform, and the derivedFrom chain.

Everything is on the device, in iOS Preferences (UserDefaults). No server, no account, no analytics, no network calls at all. Heart rate is read from HealthKit and never leaves the phone.

The flip side: delete the app and the log is gone. Export regularly — the Export tab copies a JSON envelope with every template and session, and the app nags you if it's been more than a week. Import accepts either that envelope or a bare session array, and skips malformed entries rather than rejecting the file.

On startup the app writes and reads back a probe key to verify storage actually round-trips, and shows a persistent banner if it doesn't, rather than silently losing a workout.

Branch off main, keep commits focused, open a PR. Since there's no test suite, say in the PR how you verified — browser, device, or both — and mention anything that needs a real iPhone to exercise.

If you're changing weight handling, note that stored values are always kilograms; rounding must happen in the display unit and convert back, or you'll produce weights no plate set can make.