Andromeda - Wellness Tracker (Phone + Wear OS)

Project Report

Andromeda is a modern Android wellness tracking application that helps users log daily health metrics, visualize trends, personalize their experience, and receive AI-powered wellness insights. The app is built with Jetpack Compose, follows a clean MVVM architecture with Unidirectional Data Flow (UDF), and emphasizes local-first, privacy-focused data storage.

This repository includes:

• Phone app: Home dashboard, Add logging flow, History, Settings, AI suggestions, chatbot, voice-to-text
• Wear OS companion: on-watch logging, on-watch question selection, and bidirectional sync with the phone

---

✨ Features

Daily wellness logging

• Log weight plus selected wellness metrics (diet, activity, sleep, water, protein)
• Enforces a single authoritative entry per day (same-day saves merge/update)
• Add entries for past dates via DatePicker (with careful timezone handling)
• Edit and delete entries with confirmation dialogs

Dynamic question tracking

• Select up to three wellness questions to track
• Add screen dynamically renders only the selected questions
• Unselected metrics are stored as null and do not affect UI

Home dashboard (two views)

• Swipe between:
  • Calendar view highlighting days with entries
  • Custom 14-day weight chart (Canvas-based) with dynamic scaling
• Uses only the most recent entry per day for visualization clarity
• Supports unit changes (kg ↔ lbs) consistently across UI

Personalization and settings

• Profile settings (name, age, target weight) with validation
• Light / dark theme support
• Adjustable text size
• Animated multi-layer Settings navigation

AI assistant (Gemini on Vertex AI)

• Proactive suggestions: shows exactly 3 short tips based on recent data and user profile
• Conversational chatbot with multi-turn chat history
• Voice-to-text chatbot input using Android speech recognition APIs (permission-aware)

Wear OS companion app

• Independent question selection on watch (stored locally on watch)
• Watch ↔ phone bidirectional sync:
  • Watch sends partial metrics safely (missing questions remain null)
  • Phone sends an intelligent default weight suggestion based on recent averages
  • Watch requests the default on start for a low-friction experience

Robust local persistence

• Jetpack DataStore for persistence
• JSON serialization (Gson) for complex stored objects
• Reactive updates via Kotlin Flow / StateFlow
• Full historical conversion when switching units (kg ↔ lbs), including target weight

---

✅ Features List + Status

Feature	Status
Onboarding	✔
Daily wellness logging	✔
Single entry per day enforcement	✔
Dynamic question selection (up to three)	✔
Home dashboard (calendar + 14-day chart)	✔
Calendar-based history view	✔
Edit and delete entries	✔
AI-powered suggestions (3 tips)	✔
Conversational chatbot	✔
Voice-to-text chatbot input	✔
Wear OS companion app	✔
Watch ↔ phone sync	✔
Theme & accessibility settings	✔
Local persistence (DataStore + JSON)	✔
Automatic unit conversion (kg ↔ lbs)	✔

---

🛠️ Build and Run Instructions

Prerequisites

• Android Studio (latest stable)
• Android device or emulator
  • Voice input requires a physical Android device (most emulators do not support microphone speech-to-text reliably)

Steps

1. Clone the repository

git clone https://github.com/cangokmen/CS501-Project-Andromeda.git
cd CS501-Project-Andromeda

2. Add API key

Create or edit local.properties (in the project root) and add:

GEMINI_API_KEY=your_key_here

3. Open in Android Studio

• Go to File > Open and select the project folder
• Let Gradle sync and download dependencies

4. Run the phone app

• Select an emulator or physical device
• Press Run

5. Run the Wear OS app (optional)

• Use a Wear OS emulator or a physical watch
• Run the Wear module configuration (Android Studio will show it as a separate run target)

---

🏗️ Architecture

Andromeda uses MVVM + UDF:

1. User interaction triggers a UI event
2. Event is handled by a ViewModel
3. ViewModel updates StateFlow (often via repositories)
4. UI recomposes based on the new state

UI Layer (View)

• Fully Jetpack Compose
• Single-activity structure with composable screens:
  • Home
  • Add
  • History
  • Settings
  • Chatbot
  • Login / Register (if enabled in your branch)

Navigation

• Navigation Compose
• Centralized NavHost manages screen transitions
• Routes defined with a type-safe structure (commonly a sealed Screen pattern)

ViewModel Layer

• ViewModels expose state via StateFlow
• Examples of responsibilities:
  • Home: dashboard state + suggestion loading
  • Add: dynamic form + validation + unit-safe data handling
  • Chatbot: message list, chat history, loading states
  • Watch: question selection, defaults, sync triggers

Data Layer (Model)

• Repository pattern:
  • Wellness data CRUD, date-based merge, unit conversion, sync helpers
  • User preferences (theme, text size, selected questions)
  • Profile data (name, age, target weight)
• DataStore + Gson for persistence
• Flows used as the single source of truth for reactive UI updates

---

🔒 Privacy Notes

• Designed to be local-first: wellness history and preferences are stored on-device using DataStore.
• AI features use Gemini on Vertex AI. Do not treat this project as medical software or medical advice.

---

📚 Tech Stack

• Jetpack Compose
• Navigation Compose
• ViewModel
• Kotlin Coroutines + Flow / StateFlow
• Jetpack DataStore
• Gson (JSON serialization)
• Gemini on Vertex AI (AI suggestions + chatbot)
• Android Speech APIs (voice-to-text)
• Wear OS Data Layer APIs (phone ↔ watch sync)

---

👥 Team

CAS CS 501 (Fall 2025) project by Can Gokmen and Jinpeng Huang.