Greasy Fork镜像

🚀 Major Feature: The Profile Manager Overhaul!

• 
  Complete UI Rewrite: The old right-click menu and single-profile editor have been replaced with a powerful, full-screen, two-pane Profile Manager.
  
• 
  All-in-One Interface: A new grid icon in the dock opens the manager, where you can see all your profiles in a list on the left and edit them instantly in the pane on the right.
  
• 
  Seamless Workflow: Add, edit, delete, and set default profiles all from one cohesive and intuitive screen without opening and closing multiple windows.
  

✨ New Features & Improvements

• 
  Dedicated AI Configuration Modal: The "AI Configuration" section has been moved out of the editor and into its own clean, tabbed modal. This makes managing your API key and custom refiner prompts much easier.
  
• 
  Tabbed Refiner Prompts: Easily switch between editing your refiner prompts for "with context" and "without context" using the new tab system in the AI Configuration.
  
• 
  Improved Editor Layout: The profile editor pane has been redesigned for clarity, removing clunky accordions for a more spacious and focused editing experience.
  
• 
  In-Editor Controls: You can now set a profile as default or delete it using dedicated buttons directly within the new editor pane.
  

🔧 Behind the Scenes

• Significant code refactoring for better performance, stability, and to support the new user interface.

feat(profile): add import/export and refiner model selection

Implement import and export functionality for all user-generated content, including profiles, custom icons, and AI refiner prompts. This feature is accessible via new commands in the script menu.

The import process is designed for backward compatibility, supporting both the new comprehensive data structure and the legacy profile-only array format.

Additionally, a model selection dropdown has been added to the AI refiner UI and the main settings panel. This allows users to choose which LLM to use for prompt refinement, and the selection is saved and synchronized between both locations.

fix(spa): Fix auto-injection on initial load and navigation

The script was failing to auto-inject the default profile because it wasn't loading correctly on the initial visit and wasn't aware of in-app navigation changes. This meant it only worked on a manual page refresh.

This commit fixes the issue by:

1. Widening the @match directive to https://aistudio.google.com/* to ensure the script loads on the first visit, even after a redirect.
2. Adding a MutationObserver to detect URL changes within the Single-Page Application (SPA). This allows the script to re-run the auto-injection logic whenever the user navigates to a new chat.
3. Making the script wait for the chat UI to be ready before initializing, preventing it from running too early.

Summary of changes:

Changelog: What's new in version 1.1

• Default Profile: You can now set any profile as the "Default" (via right-click on the profile icon). This profile's instructions will automatically be injected when you first load a prompt session.
• AI Refinement History: The AI Refiner tool now supports "Undo" (‹) and "Redo" (›) functionality, allowing you to step back through previous refined prompt versions.
• Refinement Guidance: Added an optional text area to the AI Refiner panel that lets you provide specific instructions (e.g., "Make the persona more aggressive") to the LLM when refining your system prompt.
• Profile Editor Redesign: The Add/Edit Profile modal is now fully responsive, utilizing exclusive accordion sections for the "System Instructions" and "AI Configuration" to ensure all content fits comfortably on small and large screens without overlapping.
• Dock UX Improvement: The invisible hover trigger area at the bottom of the screen was significantly reduced in height for less intrusive behavior.
• Dock Control: Clicking the background of the profile dock now instantly hides it and prevents accidental re-opening.