Skip to content

CHANGELOG.md

Latest commit

 

History

History
34 lines (27 loc) · 2.41 KB

CHANGELOG.md

File metadata and controls

34 lines (27 loc) · 2.41 KB
Raw

MindFlow AI Coach - Development Changelog

[Hackathon Sprint] - 2026-02-10

🚀 Major Pivot: Mobile-First & Multi-Coach Architecture

In preparation for the RevenueCat Shipyard 2026 contest (Simon @BetterCreating brief), we executed a major architectural pivot from a web-based prototype to a native Android application.

📱 Mobile Core Transition

  • Removed Web Dependencies: Stripped out dart:js_interop and package:web from audio services (AmbientSoundService, BinauralAudioService, AdvancedHypnoticAudioService). Replaced with stubbed implementations or native-compatible code to ensure Android compilation.
  • Audio Engine Update: Migrated AmbientSoundService to use audioplayers package for native mobile audio playback.
  • Backend Strategy: Switched ApiService to "Mock Mode" by default for the MVP. This decouples the app from the local Python backend, allowing it to run standalone on Android using SharedPreferences for persistence.
  • Gemini Streaming: Validated the direct HTTP streaming implementation for the chatbot, ensuring it works without a backend proxy for this hackathon build.

🧠 Multi-Coach System (The "Simon" Feature)

Implemented the core requirement of "browsing and sharing AI coaches":

  • Data Model: Defined Coach model with nlpType, systemPromptBase, and tone.
  • Coach Gallery: Built CoachGalleryScreen to browse available personas (MindFlow, The Reframer, The Clarifier, Simon, The Visualizer).
  • Personality Injection: Updated NLPPromptBuilder to inject specific NLP instructions (Milton Model, Meta Model, VAK) based on the active coach.
  • State Management: Updated ChatProvider and GeminiService to track the active coach context.

💰 Monetization (RevenueCat)

  • Integration: Verified RevenueCatService setup with Android API keys.
  • Gating: Implemented PaywallTrigger logic in CoachGalleryScreen. Users can chat with the default coach for free, but premium personas are locked behind the Pro subscription.

🛠️ Android Environment Setup

  • Installed OpenJDK 17 via Winget.
  • Configured Android SDK command-line tools.
  • Generated Android project files via flutter create . --platforms android.
  • Fixed local.properties and AndroidManifest.xml.

📝 Next Steps

  • Final APK build and testing on device.
  • Recording the demo video showcasing the Multi-Coach system.
  • Submitting to Google Play Internal Testing.