Greasy Fork

Updated deepseek models.

        "gpt-4o": "Can solve problems with images, cost-effective.",
        "gpt-4o-mini": "Handles text-only questions, cheap option.",
        "o1": "Solves image problems with highest accuracy, but is slow and expensive.",
        "o3-mini": "Handles text-only questions, fast and cost-effective, but accuracy is not as high as o1.",
        "deepseek-reasoner": "The speed is similar to o1, but the accuracy is lower than o1. It does not support image recognition and is much cheaper than o1.",
        "deepseek-chat": "The speed is similar to 4o, and the accuracy is about the same. It does not support image recognition and is the cheapest.",
        "custom": "User-defined model. Please enter your model name below."

Update Log

Version 7.0

• License Change: Updated the license to GPL v3.
  
• New Model "o1": Added "o1" as a selectable model in the dropdown menu and logged model switch events.
• New Model "o3-mini": Added "o3-mini" as a selectable model.
  
• Detailed Model Descriptions:
  
  • GPT-4o: Solves problems with images, cost-effective.
    
  • GPT-4o-mini: Handles text-only problems, cheapest option.
    
  • o1: Best accuracy for image-based problems, but slow and expensive.
    
  • o3-mini: Handles text-only problems, faster than o1 but less accurate, good cost-performance.
    
• UI Enhancement:
  
  • Displayed a brief description of the selected model to help users choose the best option based on their problem type.
• Added API Base Configuration:
  
  • Users can now set a custom API Base URL in the GUI (default: OpenAI API).
    
  • API Base is stored in localStorage and can be updated anytime.
    
• Enhanced Progress Bar:
  
  • Step-based updates to improve user experience and prevent impatience:
  • Start Phase: Displays progress bar and sets progress to 10%.
    
  • HTML Extraction: Updates progress to 20% when HTML is successfully retrieved.
    
  • Canvas Image Capture: Updates progress to 40% if an image is detected and processed.
    
  • API Request Phase: Starts a "fake progress" timer that gradually increases progress from 40% to 95% while waiting for the API response.
    
  • Request Completion: Instantly updates progress to 100% when the response is received, then hides the progress bar.
    
  • Request Failure Handling: Stops progress, resets progress bar, and displays an error message (including the error code or reason).
    
• Improved GUI Experience:
  
  • Grouped settings into sections (API Key, API Base, Model Selection, Language, Auto Answer, and Auto Submit).
    
  • Made the control panel draggable for better user interaction.

fix this

you can change api in gui now

Synchronize the latest stable version of Github

Fixed the issue of gpt request error "400" due to format error