NutriTrack

English | 日本語

An Android app, written in Kotlin with Jetpack Compose, that turns a patient's dietary-quality data (HEIFA scores) into a personal dashboard, lets them log their eating habits, and coaches them with AI-generated tips. A separate clinician view aggregates every patient's data and asks the model to surface patterns across the cohort.

Built by Haruto Iriyama.

What it does

Screen	Purpose
Welcome / Login / Register	Patients are pre-loaded from a CSV; a user claims their account by matching ID + phone number, then sets a name and password. Session state is persisted so the app reopens on the home screen.
Food-intake questionnaire	Food categories, a "persona" that best describes the user's eating style, and meal / sleep / wake times. Answers are stored in Room and pre-filled on return.
Home	The user's total Food Quality Score with a quick explanation of what it means.
Insights	Per-category HEIFA score bars (vegetables, fruit, grains, dairy, sugar, sodium, …) against their maximum, plus a share button.
NutriCoach	Look up any fruit's nutrition facts via the FruityVice API, and generate a motivational tip from Gemini that is tailored to the user's own scores. Tips are saved and can be reviewed later.
Clinician dashboard	Behind a clinician key: average HEIFA scores by sex and a "Find data patterns" button that sends the anonymised cohort data to Gemini for a written analysis.
Settings	Account details, logout, and the entry point to the clinician view.

Architecture

• UI — Jetpack Compose with Material 3 and a bottom navigation bar; one @Composable screen per feature under ui/screens.
• State — MVVM. Each screen has a ViewModel exposing StateFlows; a single NutriTrackRepository is the only thing the view models talk to.
• Persistence — Room database with Patient, FoodIntake and NutriCoachTip entities and DAOs. The patient table is seeded from assets/CustomerData.csv on first launch.
• Networking — Retrofit + Gson for FruityVice; the Google Generative AI SDK (gemini-1.5-flash) for coaching tips and cohort analysis.
• Auth — AuthManager keeps the logged-in user in SharedPreferences; passwords are set at registration and checked at login.

app/src/main/java/com/nutritrack/
├── MainActivity.kt          NavHost and bottom bar
├── NutriTrackRepository.kt  single source of truth: Room + Retrofit + Gemini
├── AuthManager.kt           session persistence, CSV loader
├── data/                    Room database, entities, DAOs
├── network/                 FruityVice Retrofit service
├── ui/screens/              one Composable per screen
├── ui/theme/                Material 3 theme
└── viewModel/               one ViewModel per screen

Running

Requires Android Studio (Ladybug or newer) and an emulator or device on API 35.

1. Create local.properties in the project root (it is git-ignored) and add a Gemini API key:

   apiKey=YOUR_GEMINI_API_KEY

2. Open the project in Android Studio and run the app configuration.

Sample patients live in app/src/main/assets/CustomerData.csv; log in with any User_ID / PhoneNumber pair from that file.

To reach the clinician dashboard, open Settings → Clinician login and enter the demo key dollar-entry-apples. It is a fixed demo value in ClinicianLoginViewModel.kt, not a real credential — a production build would check it server-side.