1.0.3

The penalty options have been removed to avoid compatibility problems with various API providers.

1.0.2

• Shorter, more precise, and simpler native prompts.
• All the native prompt instructions are now configured as system messages instead of user messages. This change should make the models follow the instructions more closely.

1.0.1

feat(version): 1.0.1

1.0.0

feat(version): 1.0.0

1.0.0-beta2

This beta version upgrades predefined prompts like "Fix Grammar" and "Improve Writing" with additional instructions to minimize unnecessary context and symbols in the responses.

1.0.0-beta1

1.0.0 (Beta 1)

This beta release builds upon the initial beta, addressing an issue with the property field. The "provider" property, previously required, is now optional.

1.0.0-beta

1.0.0 (Beta 0)

• Folder-Based Prompts: This version introduces a more flexible approach to prompt management. Instead of relying on the plugin's predefined prompts or adding them to config.json, you can now create your own prompts directly within your markdown files. This opens up greater possibilities for customization and templating in future updates.

  • Example of a prompt file with properties:
    

• Find Synonyms: Thanks to a contribution from @lucasadelino, this predefined prompt helps you find the best synonyms for a given word, presented in a clickable menu for easy selection.
  

• Patches: The "custom" provider option now automatically displays the custom base URL field, eliminating the need to enable advanced mode.

0.7.5

chore: update version to 0.7.5

0.7.4

0.7.4

1. Added support for Cohere's new command-r-plus model, featuring a 128K context window and enhanced response quality. Learn more: Introducing Command R+.
2. Reorganized Perplexity (PPLX) models: Removed some older models and added new Sonar models.
3. Integrated Gemini Model API support: The plugin now uses the model API, enabling compatibility with future Gemini model releases without hard-coding.
4. Fixed an issue where the reset models confirmation was not functioning correctly.
5. Improved layout for the advanced configuration settings tab: Settings are now separated into sections for a clearer, more intuitive user experience.

0.7.3

0.7.3

1. New Cohere model: command-r, which is free and outperforms GPT-3.5 Turbo, LLaMA 70B Chat, and Mixtral.
2. New Anthropic's Claude 3.0 Haiku: claude-3-haiku-20240307, which is fast and affordable, with a 200K context window. It also outperforms GPT-3.5 Turbo and Gemini 1.0 Pro.