Canvas Folder Organization and AI-Powered Auto-Classification

[MathisVerstrepen]

The primary goal is to enhance user experience by providing better management of a growing number of canvases, reducing clutter, improving discoverability, and streamlining workflow. This will be achieved through both manual drag-and-drop capabilities and an innovative AI-driven auto-classification mechanism.

1. History Sidebar (Left Panel) Enhancements

• Folder Display & Management:

  • Canvases will be displayed within user-created folders.
  • A default "Uncategorized" folder will implicitly exist to house all canvases not assigned to a user-created folder. This folder cannot be deleted.
  • Collapse/Expand Folder View: Users can click on a folder header (e.g., folder name or icon) to toggle its contents between a collapsed state (only folder name visible) and an expanded state (folder contents visible). This state should ideally persist per user session.

• Manual Canvas Organization (Drag & Drop):

  • Users can drag a 'canvas' element from any location (another folder, or "Uncategorized") and drop it into an existing folder.
  • Upon successful drop, the canvas is moved to the target folder.
  • Visual feedback (e.g., highlighting the target folder, ghosting the dragged item) is crucial during the drag operation to indicate valid drop targets.
  • Users should also be able to drag a canvas out of a folder back into the "Uncategorized" section.
  • Consider if reordering canvases within a folder is part of the initial scope.

• AI-Powered Auto-Classification:

  • Trigger: Via the "More Options" button (three-dots icon) then clicking "Auto-classify" will initiate the process.
  • Scope: The auto-classification can be applied to one or multiple selected canvases.
  • LLM Input:
    • For each 'canvas': Its full content (conversation history & metadata) and its title.
    • For each 'folder': Its name and its user-defined description (if present).
  • LLM Logic: A cost-effective LLM model will analyze the canvas content/title against the folder names/descriptions to determine the most relevant folder.
  • Output & User Feedback:
    • The system will move each canvas to the folder suggested by the LLM with the highest confidence score.
    • Provide clear visual feedback to the user, such as a toast notification and visually highlighting the newly moved items briefly.
  • Performance: Implement a loading state or progress indicator for the user, especially if more than one canvases are being processed.

2. Homepage (Dashboard/Landing Page)

• Folder Display: Canvases will be displayed within user-created folders in the same way as the history sidebar.
• Collapse/Expand Folder View: Similar functionality as in the history sidebar should be available for any folder lists displayed on the homepage.

3. Settings (Folder Management)

• Create New Folder:

  • Users can create new folders via a dedicated UI section in the settings.
  • Each folder must have:
    • Name (required): A user-defined string.
    • Description (optional): A text area for users to provide a detailed description of the folder's purpose or expected content. This is critical for improving the accuracy of the LLM's auto-classification.
  • (Consider: Optional icon or color for visual distinction, but not a P0 for initial implementation).

• Edit Folder:

  • Users must be able to edit the name and description of existing folders.

• Delete Folder:

  • Users can delete existing folders.
  • Behavior on Deletion: When a folder is deleted, all canvases contained within it must be automatically moved to the "Uncategorized" section. A confirmation dialog should warn the user about this action before proceeding.

• Reorder Folders:

  • Allow users to manually reorder their folders in the settings. This custom order should be reflected consistently across the history sidebar and homepage.

4. General Considerations

• Persistence: All folder structures, canvas assignments to folders, and folder collapse/expand states must persist across user sessions (e.g., stored in user preferences or backend database).
• Empty Folders: How are empty folders displayed?
• Performance: Ensure that loading and displaying a potentially large number of folders and canvases, especially after auto-classification, remains performant and responsive.