…modes

- Move execution controls to right side of toolbar with spacer
- Fix radio buttons to always use circular indicators
- Improve text contrast and readability for execution mode selection
- Add proper hover states and consistent sizing for radio buttons

- Add UI update throttling to prevent redundant live mode updates
- Cache environment validation results to avoid repeated file system checks
- Optimize event handler setup to be incremental, not full rebuild
- Reduce logging spam during live mode activation
- Keep event handlers persistent for faster mode switching
- Fix performance regression introduced by environment management system

- Change button text color to black for better visibility on bright backgrounds
- Remove emoji icons from all execution button text states
- Keep white text only on red (paused) button for contrast
- Update all button states: Execute Graph, Start Live Mode, Pause/Resume, etc.

- Changed 'Run' menu to 'Environment' menu since it only contains environment management
- More descriptive and intuitive menu organization
- Better reflects the actual functionality contained within

…pacing

- Add graph description extraction from markdown H1 titles and following paragraphs
- Implement Graph Properties dialog accessible via File → Graph Properties menu
- Fix paragraph spacing preservation in both node and graph descriptions
- Update FlowFormatHandler to properly parse and generate graph metadata
- Ensure copy/paste operations exclude graph descriptions as requested
- Add proper paragraph break handling with double newlines between paragraphs
- Rename json_to_flow/flow_to_json methods to data_to_markdown/markdown_to_data
- Include actual node descriptions in saved markdown instead of placeholder text

🤖 Generated with [Claude Code](https://claude.ai/code)

- Add description field to Node class with serialization support
- Enhance Node Properties dialog with description editor (QTextEdit)
- Update dialog constructor to accept and display node descriptions
- Preserve node descriptions in save/load cycle
- Support markdown formatting in node description editor

🤖 Generated with [Claude Code](https://claude.ai/code)

- Remove JSON file format from save/load dialogs (only .md files supported)
- Update _save_file() to use actual graph title/description instead of defaults
- Simplify _load_file() to only handle .md format using FlowFormatHandler
- Use data_to_markdown() and markdown_to_data() method names consistently
- Ensure graph metadata is properly preserved during save/load operations

🤖 Generated with [Claude Code](https://claude.ai/code)

- Update save_view_state() to accept optional file_path parameter
- Simplify view state loading logic with robust transform and center point handling
- Add comprehensive GUI-based test for view state persistence functionality
- Test covers zoom/pan operations, file switching, and state restoration
- Ensure view state is properly saved when switching between files

🤖 Generated with [Claude Code](https://claude.ai/code)

- Add detailed graph description explaining the complete workflow
- Enhance node descriptions with multi-paragraph technical details
- Include implementation specifics for each component
- Demonstrate proper paragraph spacing in descriptions
- Update node positions and GUI state after testing

🤖 Generated with [Claude Code](https://claude.ai/code)

- Fixed startup venv dialog not showing for new/example graphs
- Added proper environment selection prompts before execution
- Fixed QSettings path errors in environment manager using correct project root detection
- Fixed view state (pan/zoom) not being restored when loading last saved graph on startup
- Set default environment to True for new/untitled graphs
- Added ensure_environment_selected() method called before execution

Two major fixes for the pin generation system:

1. Fix _parse_type_hint() method to properly handle complex generic types like List[Dict]:
   - Added support for ast.Name slices (simple generics)
   - Added support for ast.Tuple slices (multiple parameters)
   - Added fallback handling for other slice structures

2. Fix update_pins_from_code() logic error for List/Tuple detection:
   - Changed to only treat Tuple types as multiple outputs
   - Handle List[Dict] and other single-element generics correctly
   - Prevents AttributeError when accessing .elts on non-Tuple slices

Example graph improvements:
- Fixed missing GUI State Handler functions (set_values, set_initial_state)
- Connected data-generator to rest of analysis pipeline
- Removed non-functional buttons from dashboard and password generator
- Updated node positions and metadata from GUI testing

🤖 Generated with [Claude Code](https://claude.ai/code)

- Add missing workflow starter entry point node for clean initiation
- Fix broken data flow connections between all nodes in pipeline
- Refactor complex @node_entry functions following FlowSpec best practices:
  * File Categorizer: Extract FileCategorizer class with modular methods
  * File Mover: Extract FileOrganizer class with specialized operations
  * Operation Verifier: Extract OrganizationVerifier class with focused methods
- Add comprehensive error handling and validation throughout workflow
- Enhance GUI elements with tooltips and better user experience
- Fix folder scanner to return tuple with both file list and base path
- Add operation verification node for post-organization validation
- Create complete data pipeline: Scanner → Categorizer → Creator → Mover → Verifier
- Ensure all pin connections properly match function signatures
- Improve code maintainability and testability with helper classes

🤖 Generated with [Claude Code](https://claude.ai/code)