Skip to content

CHANGELOG.md

Latest commit

 

History

History
247 lines (201 loc) · 8.39 KB

CHANGELOG.md

File metadata and controls

247 lines (201 loc) · 8.39 KB
Raw

Changelog - D.A.V.I.D AI

All notable changes to this project will be documented in this file.

[Unreleased] - 2026-01-13

Added

  • WeatherService: New weather integration with Open-Meteo API

    • Real-time weather data with actual temperature, conditions, humidity, and wind speed
    • Location-based weather queries (e.g., "weather in New York")
    • Natural language weather descriptions with contextual advice
    • No API key required - uses free Open-Meteo API
    • Geocoding support for city names
    • Default location fallback (Kolkata)

  • Male Voice Support: Added David (male) voice option

    • Voice selection between David (male) and Dayana (female)
    • Automatic pitch adjustment for male/female voices
    • Gender-specific voice detection from system TTS
    • Voice preferences saved in settings

  • Code Filtering: Comprehensive filtering system

    • Removes code blocks, JSON, arrays, and technical syntax from responses
    • Filters debug logs, status emojis, and internal tags
    • Prevents TTS from speaking code or technical terms
    • Cleans up chat responses for natural conversation

Fixed

Chat System

  • Weather Integration: Chat now returns actual weather data instead of web URLs

    • Integrated WeatherService for real-time weather information
    • Removed web search for weather queries
    • Weather responses include temperature in both Celsius and Fahrenheit
    • Contextual weather advice (umbrella reminders, heat warnings, etc.)

  • Response Filtering: Fixed chat responses containing internal code

    • All responses now filtered before sending to user
    • Removed code patterns, debug logs, and technical syntax
    • Cleaner, more natural conversation flow

  • Model Loading: Improved LLM model detection and loading

    • Better error handling for model initialization
    • Fallback to smart responses when model not available
    • Status indicators show model availability

Gesture Recognition

  • Model Loading: Fixed gesture models not loading from ModelManager

    • Properly checks ModelManager's david_models directory
    • Detects downloaded .task and .tflite gesture model files
    • Validates model file size (minimum 1MB)
    • Clear error messages when models not found
    • Status shows whether models are loaded or need downloading

  • Initialization: Improved MediaPipe initialization

    • Better error logging with stack traces
    • Separate initialization for hand detection and gesture recognition
    • Fallback mode when models not available
    • Status indicators: "✅ Models loaded" vs "❌ Models not loaded"

  • Model Detection: Enhanced model file detection

    • Looks for files containing "hand", "gesture", "palm"
    • Supports both .task (MediaPipe) and .tflite (TensorFlow Lite) formats
    • Logs available files for debugging
    • Validates model integrity before loading

Voice System

  • Male Voice: Added support for male voice (David)

    • getAvailableVoices() now returns both David (Male) and Dayana (Female)
    • Voice selection properly saved in SharedPreferences
    • Automatic pitch adjustment (0.8 for male, 1.2 for female)
    • Searches system TTS for gender-specific voices
    • Falls back to pitch adjustment if specific voice not found

  • Code Speech: Fixed TTS speaking internal code

    • filterInternalCode() function filters text before speech
    • Removes code blocks, JSON, arrays, HTML/XML tags
    • Strips debug logs, status emojis, technical syntax
    • Prevents speaking of variable declarations, imports, packages
    • Cleans emoji characters that shouldn't be vocalized

  • Voice Management: Improved voice selection

    • changeVoice() properly updates preferences
    • getCurrentVoice() retrieves saved selection
    • Better logging of available system voices
    • Locale-aware voice selection (English preference)

Improved

Model Manager

  • Already functioning correctly for model downloads
  • Models properly saved to app/files/david_models/
  • Download progress tracking with pause/resume support
  • Model validation and integrity checks
  • Essential models detected based on device RAM

Chat Engine

  • Enhanced fallback responses with Nexuzy Tech branding
  • Better handling of greetings, questions, and commands
  • Math calculation support
  • Time and date queries
  • General knowledge responses
  • Jokes and personality

Error Handling

  • Better error messages throughout the app
  • Detailed logging for debugging
  • User-friendly error descriptions
  • Graceful fallbacks when features unavailable

Technical Details

New Files

  • app/src/main/kotlin/com/davidstudioz/david/weather/WeatherService.kt
    • Complete weather service implementation
    • Open-Meteo API integration
    • Geocoding and location services
    • Weather condition mapping (WMO codes)

Modified Files

  • app/src/main/kotlin/com/davidstudioz/david/chat/ChatEngine.kt

    • Added WeatherService integration
    • Added code filtering for responses
    • Improved weather query detection
    • Enhanced model status reporting

  • app/src/main/kotlin/com/davidstudioz/david/gesture/GestureController.kt

    • Fixed model directory path (uses ModelManager's directory)
    • Enhanced model file detection logic
    • Better initialization error handling
    • Improved status reporting
    • Added model validation checks

  • app/src/main/kotlin/com/davidstudioz/david/voice/TextToSpeechEngine.kt

    • Added male voice support (David)
    • Added code filtering before speech
    • Added getAvailableVoices() method
    • Improved voice selection logic
    • Added gender-based pitch adjustment
    • Added isSpeaking() status check

Dependencies

  • No new dependencies added
  • Uses existing libraries:
    • MediaPipe for gesture recognition
    • Android TTS for speech
    • OkHttp for network requests
    • Gson for JSON parsing

Breaking Changes

  • None - all changes are backward compatible

Migration Notes

  • Users may need to re-download gesture models if previously downloaded incorrectly
  • Voice preferences will be migrated automatically
  • Weather data will now show actual information instead of web links

Known Issues

  • Gesture recognition requires camera permission
  • Male voice quality depends on system TTS voices available
  • Weather service requires internet connection
  • Location permission needed for automatic location detection

Credits

Testing Checklist

Weather Service

  • Test "what's the weather" query
  • Test "weather in [city]" query
  • Test weather with no internet
  • Verify temperature in Celsius and Fahrenheit
  • Check contextual advice (rain, snow, heat, cold)

Chat System

  • Verify weather returns actual data, not URLs
  • Check responses don't contain code
  • Test with LLM model loaded
  • Test without LLM model (fallback)
  • Verify model status display

Gesture Recognition

  • Download gesture models from settings
  • Check model status shows "loaded" or "not loaded"
  • Start gesture detection
  • Verify no "model not loaded" error when models exist
  • Test hand tracking
  • Test gesture recognition

Voice System

  • Check voice settings show both David (Male) and Dayana (Female)
  • Select David voice and verify lower pitch
  • Select Dayana voice and verify higher pitch
  • Test TTS doesn't speak code or emojis
  • Verify voice preference saves correctly

Integration

  • Test full conversation flow
  • Ask for weather via voice
  • Verify gesture models load on app start
  • Check all features work together

Future Improvements

Planned Features

  • Weather forecast (next 7 days)
  • Weather alerts and notifications
  • More gesture types (swipe, pinch, rotate)
  • Custom voice training
  • Offline weather caching
  • Weather widget
  • Voice emotion detection
  • Gesture customization

Performance Optimizations

  • Lazy loading for models
  • Background model preloading
  • Weather data caching
  • TTS response queue optimization
  • Gesture processing on GPU

UI/UX Enhancements

  • Voice preview in settings
  • Gesture training mode
  • Weather visualization
  • Model download progress in status bar
  • Interactive gesture tutorial

Support

For issues, feature requests, or questions:

License

Copyright © 2026 Nexuzy Tech Ltd. All rights reserved.