implement training report

[gbeane]

Add Training Report Dialog & Window Menu

Summary

Replaces terminal output during classifier training with a non-modal dialog that displays training performance metrics. Training reports are also saved as Markdown files in the project for documentation and comparison purposes. Additionally, adds a Window menu for improved window management across JABS.

Changes

New Features

• Training Report Dialog (src/jabs/ui/training_report.py)

  • Non-modal dialog using QWebEngineView to display HTML-formatted training reports
  • Can be reused across training sessions, preserving window position and size
  • Copy to Clipboard button to copy Markdown report
  • Shows informational text about report save location
  • Deferred HTML content loading to prevent "Compositor returned null texture" errors

• Training Report Generation (src/jabs/classifier/training_report.py)

  • TrainingReportData dataclass for structured training information
  • CrossValidationResult dataclass for individual CV iteration results
  • generate_markdown_report() - Creates Markdown-formatted reports with timestamp
  • save_training_report() - Saves reports to disk

• Window Menu (src/jabs/ui/main_window.py)

  • Standard macOS-style Window menu with:
    • Minimize (⌘M / Ctrl+M)
    • Zoom (toggle maximize/restore)
    • Bring All to Front
  • Dynamic window list showing all open JABS windows:
    • Main Window
    • User Guide (when open)
    • Training Report dialogs (when open)
  • Checkmark (✓) indicates currently active window
  • Click any window to activate and bring to front

Modified Components

• TrainingThread (src/jabs/ui/training_thread.py)

  • Removed all print() statements
  • Added training_report signal that emits Markdown content
  • Generates and saves training reports to jabs/training_logs/
  • Report filenames include behavior name and timestamp (e.g., Grooming_20260102_143022_training_report.md)

• CentralWidget (src/jabs/ui/central_widget.py)

  • Connected training_report signal to new _on_training_report() slot
  • Modified _training_thread_complete() to display or update training report dialog
  • Reuses existing dialog when multiple trainings are run
  • Recreates dialog when JABS has focus to ensure proper focus switching
  • Quietly updates existing dialog when JABS doesn't have focus (no focus stealing from other apps)
  • Added get_open_dialogs() public method for Window menu integration
  • Stores Markdown temporarily between report generation and training completion

• TrainingReportDialog (src/jabs/ui/training_report.py)

  • HTML rendering moved into dialog class as _markdown_to_html() method
  • Added update_content() method to refresh dialog with new training results
  • Proper encapsulation with UI-specific logic contained in UI module

• UserGuideDialog (src/jabs/ui/user_guide_dialog.py)

  • Removed WindowType.Tool flag to prevent staying on top of other JABS windows
  • Deferred initial content loading to prevent compositor texture errors
  • Now appears in Window menu when open

Documentation Updates

• gui.md

  • Updated Train Button description to mention training report dialog
  • Added new "Training Reports" section explaining report contents and use cases
  • Added Window menu documentation explaining all menu items and dynamic window list

• keyboard-shortcuts.md

  • Replaced all occurrences of Cmd+ with ⌘ symbol

• project-setup.md

  • Added jabs/training_logs directory documentation

Tests

• New Test Files:
  • tests/classifier/test_training_report.py - 18 tests for report generation and saving
  • tests/ui/test_training_report_dialog.py - 13 tests for dialog UI components

Training Report Contents

Each training report includes:

• Training Summary: Behavior name, classifier type, window size, distance unit, training time, timestamp
• Label Counts: Frame and bout counts for both classes
• Cross-Validation Results: Performance metrics (accuracy, precision, recall, F1) for each iteration with test video/identity
• Feature Importance: Top 20 features from the final trained model

Window Management Improvements

The new Window menu provides:

• ✅ Discoverability: Users can see all open JABS windows in one place
• ✅ Keyboard accessible: All window management via keyboard navigation
• ✅ Smart focus handling: Training report only steals focus when user is working in JABS
• ✅ Reusable dialogs: Training report dialog preserves position/size across training sessions

Benefits

• ✅ Better UX: Reports appear automatically in an easy-to-read format
• ✅ Permanent Records: All training sessions are logged to disk with timestamps
• ✅ Comparison: Users can compare different classifier configurations over time
• ✅ Documentation: Training reports serve as analysis workflow documentation
• ✅ Improved Window Management: Easy switching between JABS windows via Window menu
• ✅ Respects User Focus: Doesn't steal focus from other applications
• ✅ Persistent Dialogs: Training report and User Guide remain accessible across sessions

Screenshots

Testing

• ✅ Tested with cross-validation (k > 0) and without (k = 0)
• ✅ Verified Markdown files are saved correctly to jabs/training_logs/
• ✅ Confirmed HTML rendering displays all tables and metrics correctly
• ✅ Validated dialog appears after training and can be reused
• ✅ Validated no output printed to Terminal during training
• ✅ Tested Window menu with multiple open windows
• ✅ Verified focus behavior when JABS has/doesn't have focus
• ✅ Tested dialog reuse preserves position and size
• ✅ Verified User Guide no longer stays on top of other JABS windows

Linked Issues

Closes #140

[Copilot]

Pull request overview

This PR replaces terminal output during classifier training with a modal dialog that displays formatted training reports. Reports are automatically saved as Markdown files in the project's jabs/training_logs/ directory for documentation and comparison purposes.

Key changes:

• New training report dialog that displays HTML-formatted performance metrics
• Training report generation system that creates Markdown and HTML reports
• Removal of all print statements from the training process

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
src/jabs/ui/training_thread.py	Removed print statements; added report generation and HTML emission
src/jabs/ui/training_report.py	New dialog component for displaying training reports
src/jabs/ui/central_widget.py	Connected training report signal and displays dialog after training
src/jabs/classifier/training_report.py	New module with report generation functions and data structures
src/jabs/classifier/classifier.py	Refactored feature importance into reusable getter method
src/jabs/classifier/init.py	Exported new training report classes and functions
src/jabs/project/project_paths.py	Added training_log_dir property and directory creation
src/jabs/project/session_tracker.py	Removed unused fbeta_notbehavior parameter
src/jabs/resources/docs/user_guide/gui.md	Documented new training report feature
src/jabs/resources/docs/user_guide/project-setup.md	Documented jabs/training_logs directory

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 11 out of 11 changed files in this pull request and generated 1 comment.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 16 out of 16 changed files in this pull request and generated 1 comment.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.