Ayuvo

Your whole health, in one place.

Website · Contribute · Security · MIT licence

Private, open-source, all-in-one health companion for iPhone (SwiftUI) and Android (Jetpack Compose): nutrition, workouts, health records, medications, a local mirror of Apple Health / Health Connect, fasting, water and an AI coach that runs on the API key the user brings, or on a model that stays on the phone. No account, no Ayuvo servers, no analytics.

Inspired by Fud AI by Apoorv Darshan. Ayuvo started from that MIT-licensed calorie tracker; its copyright notice is kept in LICENSE and shown in the apps under Settings → Legal → Licenses.

What Ayuvo is

Area	What it does
Nutrition	Meals by photo (up to 10 at once), barcode (Open Food Facts), voice, text, saved meals or manual entry; review calories, macros and 30+ nutrients before saving; "What if?" previews; personalised targets (Mifflin-St Jeor / Katch-McArdle); custom meal times; nutrient goals and Home cards.
Workouts	Day-by-day diary with sets, reps, weight, RPE; calculated burn; 1,300-exercise library (shared/exercises) with photos/animations streamed from the upstream dataset.
Health data hub	Every Health Connect / HealthKit type the user grants, all history, mirrored into a local SQLite database (shared/health/schema.sql, metric_registry.json); Home tiles, category lists, D/W/M/6M/Y charts, Show All Data, Data Sources, unit options; export/import as ayuvo-health-data zips; never in cloud backups. Contract: docs/health-data.md.
Camera vitals	Manual finger scan (rear camera + torch, PPG) and face scan (front camera, remote PPG): heart rate, pulse intervals, HRV, respiratory rate, finger-vs-face compare, personal baselines and a recovery indicator. SpO₂ (experimental) and blood pressure (research) only after the user turns them on and calibrates against their own oximeter or cuff. Every value carries a classification and a confidence. Frames are processed on the device and never stored; scans are not written to Health. General wellness, not a medical measurement. Contract: docs/camera-vitals.md.
Cycle tracking	Opt-in period tracker: one-tap period start/end, calendar, daily flow, pain, symptoms, mood and notes, history, trend charts, discreet reminders, and calendar-based estimates of the next period, fertile window and ovulation (labelled as estimates; not birth control, never a diagnosis). Stored in its own local database excluded from cloud backups; optional read/write with Apple Health / Health Connect; Coach access only after consent. Contract: docs/cycle-tracking.md.
Fasting & water	Optional fasting timer (1–168 h) with editable history; optional water goals, reminders, widgets. Neither is written to Health.
AI coach	Multi-turn chat with tool calls over the user's own data (diary, workouts, weight trend, fasts, and Health data only behind a visible consent toggle). 15 BYOK providers, or on-device models (Gemma 4 E2B, Qwen3 and Google's MedGemma 1.5 4B, a gated research model under the Health AI Developer Foundations terms, not a medical device) / Apple Intelligence; see local-models/README.md; keys in Keychain / EncryptedSharedPreferences.
Platform	Apple Watch app + complications, iOS/Android widgets, Siri Shortcuts, iOS Share Extension, 18 languages, 18 accent tints with matching icons, Export/Import All Data zip on both platforms, which also moves a profile and its logs between iPhone and Android (docs/portable-data.md), plus optional Google Drive backup on Android (the Drive backup never includes health data or coach chat).

Tabs on both platforms: Home · Health · Coach · Workouts · Settings. The Health tab switches between Progress (weight, body fat, calories, workouts) and Health Data (the hub).

Privacy model

Everything is stored on the device. Data leaves only when the user acts, and only to a service they chose:

Flow	Sent	To	When
BYOK AI	photos, text, coach messages + requested context (health data only with the toggle)	provider the user configured	analyse / ask
Speech-to-text	audio clip	provider the user chose (or on-device)	voice input
Barcode	barcode number	Open Food Facts	scan
Exercise media	plain GET	raw.githubusercontent.com (© Gym visual)	browsing Workouts
Model download	plain GET	Hugging Face	tapping Download
Cloud backup (Android)	archive without health DB / chat	user's own Google Drive	toggle on
Google Health	OAuth sign-in; read-only health data downloaded	Google (accounts.google.com, oauth2.googleapis.com, openidconnect.googleapis.com, health.googleapis.com); Android custom-client mode also opens a one-shot 127.0.0.1 listener during consent	connect, Sync now, app open if auto-sync on
Update check / review	none (public metadata)	Apple / Google	app open
OS backups	per OS settings; health DB and chat excluded	Apple / Google	—

No analytics, crash or ad SDKs; no accounts; no first-party endpoints. Copy rule: never claim data "never leaves the device"; say "kept on this device, never stored in Drive backup, shared with your AI provider only through Coach".

Repo layout

ayuvo/
├── ios/            # SwiftUI app (target calorietracker), widgets, watch app, share extension, tests
├── android/        # Kotlin + Compose app (com.ayuvo.health), tests, docs
├── shared/         # exercises dataset + health registry/schema shared by both apps; shared/vitals = camera vitals DSP contract; shared/cycle = period tracker contract
├── web/            # static site (Firebase Hosting): landing, feature pages, privacy, terms, support. Generated by scripts/web_build.py
├── brand/          # mark geometry outputs, tints, fonts, brand README
├── scripts/brand/  # icon/marketing renderer (Pillow), scripts/web_check.py, scripts/brand_residue.py
├── scripts/web/    # site content + components, screenshot redaction, social cards (scripts/web_build.py builds web/)
├── marketing/      # logos, store graphics, screenshot storyboard + seed guide
├── docs/           # health-data contract, export format, cloud backup
├── local-models/   # on-device model catalog + third-party notices
├── APPSTORE.md · PLAYSTORE.md · RELEASE_NOTES.md · DEVELOPMENT.md · SECURITY.md
└── LICENSE · THIRD_PARTY_NOTICES.md · ASSET_CREDITS.md

Build & run

# iOS
xcodebuild -project ios/calorietracker.xcodeproj -scheme calorietracker \
  -destination 'platform=iOS Simulator,name=iPhone 17 Pro' build

# Android
export ANDROID_HOME=~/Library/Android/sdk
cd android && ./gradlew :app:assembleDebug
adb install -r app/build/outputs/apk/debug/app-debug.apk
adb shell am start -n com.ayuvo.health.debug/com.ayuvo.health.MainActivity

First launch: onboarding (profile, goals, notifications, Apple Health / Health Connect with the Coach consent toggle, AI provider key). See DEVELOPMENT.md for tests, the web site, the brand pipeline, screenshots and the release checklist.

Release process

• Versions: iOS MARKETING_VERSION / CURRENT_PROJECT_VERSION in ios/calorietracker.xcodeproj/project.pbxproj; Android versionName / versionCode in android/app/build.gradle.kts.
• Notes: add a ## vX.Y section to RELEASE_NOTES.md (scripts/release_notes.py refuses to publish without it).
• Tags: vX.Y → Xcode Cloud archive + GitHub release entry; android-vX.Y → signed AAB uploaded to Play as a draft.
• Store copy: APPSTORE.md, PLAYSTORE.md; site: cd web && firebase deploy --only hosting.

Contributing

Ayuvo is open source at github.com/astechtic/ayuvo. Code, translations, exercise data and docs are welcome: see CONTRIBUTING.md. Report vulnerabilities privately as described in SECURITY.md.

Licence

MIT — see LICENSE. Ayuvo is inspired by, and built on, the MIT-licensed Fud AI code base; the original copyright notice is retained there and shown in the apps under Settings → Legal → Licenses. Third-party credits: THIRD_PARTY_NOTICES.md, ASSET_CREDITS.md, local-models/legal/.