Impend - Privacy-First AI Behavioral Finance Coach

Mission

Impend is a production-grade, globally scalable, privacy-first AI-powered behavioral finance coaching platform. Built using Kotlin Multiplatform (KMP) with strict MVVM architecture, it ensures all sensitive financial data stays entirely on-device.

Non-Negotiable Principles

• 100% On-Device Financial Data: No raw expense data ever leaves the device.
• Strict MVVM: UI -> ViewModel -> Domain <- Data.
• Clean Architecture: Repository pattern is mandatory.
• Shared Business Logic: The KMP shared module contains all core business rules and analytics.
• No Ads, No Bank Scraping: Privacy is our core feature.

---

Architectural Overview

graph TD
subgraph UI Layer [Compose Multiplatform / Web]
AppUI[App UI/Screens]
end
subgraph Presentation Layer [Shared KMP]
VM[Shared ViewModels]
State[StateFlow / UI State]
end
subgraph Domain Layer [Shared KMP]
UC[Use Cases]
Repo[Repository Interfaces]
Models[Domain Models / Entities]
end
subgraph Data Layer [Shared KMP]
SQL[SQLDelight Local DB]
Prefs[Encrypted Storage]
Ktor[Ktor Network API - NO FINANCE DATA]
end
subgraph AI Analytics Engine [Shared KMP]
Stats[Time-Series & Z-Score Engine]
Scores[Impulse Risk Score™]
end
%% Dependencies
AppUI --> VM
VM --> State
State --> AppUI
VM --> UC
UC --> Repo
UC --> Stats
Stats --> Repo
DataImpl[Repository Implementations] -.-> Repo
DataImpl --> SQL
DataImpl --> Prefs
DataImpl --> Ktor

Loading

Module Structure

• :composeApp
  • /commonMain: Shared compose UI, screen navigations.
  • /androidMain: Android specific entry point.
  • /iosMain: iOS specific entry point.
• :shared
  • /commonMain: All viewmodels, domain models, algorithms, and data access.
  • /androidMain: Android specific driver/sql/platform dependencies.
  • /iosMain: iOS specific driver/sql/platform dependencies.

Key Features

Free Tier

• Manual expense entry & Category tagging
• Mood logging (1-5 scale)
• Weekly insights (max 3) & Basic charts
• Global Context: Multi-currency support (USD, EUR, GBP, JPY, INR) and multi-lingual UI (English, Spanish).

Pro Tier (Subscription)

• Unlimited insights
• Impulse Risk Score™ algorithm
• Behavioral clustering & Anomaly detection
• PDF export

---

CI/CD & Distribution

Android Release Pipeline

The project includes a robust GitHub Actions workflow for automated Android App Bundle (AAB) distribution to the Google Play Store.

• Workflow: .github/workflows/android_release.yml
• Triggers: On push to release/* branches or manual dispatch.
• Process:
  • Builds signed AAB using Java 17.
  • Automatically uploads to the Internal Testing track on Google Play Console.

Required GitHub Secrets

To enable the release pipeline, configure the following secrets in your repository:

• KEYSTORE_BASE64: Base64 encoded .jks keystore file.
• KEYSTORE_PASSWORD: Password for the keystore.
• KEY_ALIAS: Alias for the signing key.
• KEY_PASSWORD: Password for the signing key.
• PLAY_CONSOLE_JSON: Service account JSON for Google Play API access.

---

Getting Started

Prerequisites

• Android Studio Ladybug or newer.
• JDK 17.
• Xcode (for iOS builds).

Running the App

1. Clone the repository.
2. Open in Android Studio.
3. Sync Gradle.
4. Run the :composeApp module on an emulator or physical device.

---

Testing

Run shared module tests using:

./gradlew :shared:allTests