Voice command improvements:
- Auto-start listening when voice dialog opens (removes extra tap/click)
- Added Activity option to unknown intent dialog
- Users can now speak immediately after clicking the mic button
Desktop navigation:
- Added Home icon button in top bar/header for quick navigation to main page
- Positioned between app title and user avatar
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Added new tracking pages:
- Medicine tracker: track medication name, dosage, unit, route, and reason
- Activity tracker: track play, exercise, walks, music, reading, tummy time, etc.
- Both pages follow existing tracker patterns with recent activities list
Voice command improvements:
- Updated voice classification to support medicine and activity types
- Added detailed extraction fields for medicine (medicineName, dosage, unit, route, reason)
- Added detailed extraction fields for activity (activityType, duration, description)
- Enhanced unknown intent dialog with manual tracker selection
- Updated tracker options to match implemented pages (removed milestone)
Backend changes:
- Added MEDICINE and ACTIVITY to ActivityType enum
- Created migration V013 to add medicine/activity to database CHECK constraint
- Updated voice service prompts to include medicine and activity extraction
Frontend changes:
- Created /track/medicine page with full CRUD operations
- Created /track/activity page with full CRUD operations
- Added Medicine card to /track page with MedicalServices icon
- Updated VoiceFloatingButton unknown dialog with 4 tracker options
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Fixed multiple issues with voice command workflow:
**Status Transition Fixes:**
- Fixed infinite loop in status update useEffect by checking if status actually needs to change
- Status now properly transitions: listening → understanding → review/close
- Added debug logging to track status changes
**UI Bug Fixes:**
- Fixed crash in diaper tracker when conditions field is undefined (voice-created activities)
- Auto-close dialog when classification returns "unknown" type
- Added optional chaining for conditions.join() in getDiaperDetails
**Changes:**
- VoiceFloatingButton: Prevent setting same status repeatedly
- VoiceFloatingButton: Close dialog on unknown classification
- Diaper page: Handle missing conditions field gracefully
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Implemented complete review/edit workflow for voice commands with ML feedback collection:
**Backend:**
- Created V012 migration for voice_feedback table with user action tracking
- Added VoiceFeedback entity with approval/edit/reject actions
- Implemented voice feedback API endpoint (POST /api/v1/voice/feedback)
- Fixed user ID extraction bug (req.user.userId vs req.user.sub)
**Frontend:**
- Built VoiceActivityReview component with field-specific editors
- Integrated review dialog into voice command workflow
- Added approve/edit/reject handlers with feedback submission
- Fixed infinite loop by tracking processed classification IDs
**Features:**
- Users can review AI-extracted data before saving
- Quick-edit capabilities for all activity fields
- Feedback data stored for ML model improvement
- Activity creation only happens after user approval/edit
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Backend changes:
- Update LLM prompt to use correct field names matching frontend interfaces
- Use 'diaperType' instead of 'type' for diaper activities
- Use 'feedingType' instead of 'method' for feeding activities
- Simplify sleep structure (duration, quality, location only)
Frontend changes:
- Add processedClassificationId tracking to prevent infinite loop
- Create unique ID for each classification to avoid duplicate processing
- Reset processed ID when dialog opens/closes or new recording starts
This fixes the issue where voice commands created multiple duplicate
activities and had mismatched data structures causing tracker warnings.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Fix login endpoint to return families as array of objects instead of strings
- Update auth interface to match /auth/me endpoint structure
- Add silence detection to voice input (auto-stop after 1.5s)
- Add comprehensive status messages to voice modal (Listening, Understanding, Saving)
- Unify voice input flow to use MediaRecorder + backend for all platforms
- Add null checks to prevent tracking page crashes from invalid data
- Wait for auth completion before loading family data in HomePage
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Replace navigation to pre-filled forms with direct API activity creation
- Fetch children from family and use first child (can be enhanced for name matching)
- Show success/error messages with proper feedback
- Auto-close dialog after successful save
- Add test endpoint /api/v1/voice/test-classify for easy testing
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add URL parameter reading to diaper tracking page for voice-extracted data
- Add comprehensive server-side logging in voice controller and service
- Log request type (Web Speech API vs MediaRecorder), input text/audio, GPT calls, and classification results
- Enable automatic form pre-filling when voice commands navigate to tracking pages
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Set continuous=true to keep listening through pauses
- Only process final results, ignore interim transcripts
- Add usesFallback check to route Web Speech API transcripts through classification
- Desktop now captures complete phrases before classification
- Add detailed logging for debugging recognition flow
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Remove temperature parameter from GPT-5-mini activity extraction (not supported)
- Add classification state to useVoiceInput hook to avoid duplicate API calls
- Prevent infinite loop in VoiceFloatingButton by tracking lastClassifiedTranscript
- Use classification from backend directly instead of making second request
- iOS Safari now successfully transcribes with Azure Whisper and classifies with GPT-5-mini
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Force MediaRecorder fallback for all iOS Safari devices
- Add iOS device detection to avoid Web Speech API on iOS
- Support multiple audio formats (webm, mp4, default) for compatibility
- Add comprehensive error logging throughout the flow
- Improve error messages with specific guidance for each error type
- Add console logging to track microphone permissions and recording state
- Better handling of getUserMedia permissions
This should help diagnose and fix the "Failed to recognize speech" error
by ensuring iOS Safari uses the MediaRecorder path with proper permissions.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Frontend changes:
- Add MediaRecorder fallback for iOS Safari (no Web Speech API support)
- Automatically detect browser capabilities and use appropriate method
- Add usesFallback flag to track which method is being used
- Update UI to show "Recording..." vs "Listening..." based on method
- Add iOS-specific indicator text
- Handle microphone permissions and errors properly
Backend changes:
- Update /api/v1/voice/transcribe to accept both audio files and text
- Support text-based classification (from Web Speech API)
- Support audio file transcription + classification (from MediaRecorder)
- Return unified response format with transcript and classification
How it works:
- Chrome/Edge: Uses Web Speech API for realtime transcription
- iOS Safari: Records audio with MediaRecorder, sends to server for transcription
- Fallback is transparent to the user with appropriate UI feedback
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
