v4.3

This update brings new providers, large performance improvements, and various fixes.

What's Changed

• Add AI4Chat (gpt-4) and DarkAI (gpt-4o) providers
• Made a significant optimization in AI Chat, which is especially noticeable with a large number of chats.
• Made the Next Chat and Previous Chat actions consistent with the chats order in the dropdown list
• Fixed multiple providers, along with other bug fixes
• Fixed image generation and added more image models
• Improved error handling in AI Chat, allowing for partial recovery of data (experimental)
• Several (non-functional) codebase refactors

Known issues

• Web Search functionality is not working for many users. A fix is in progress, and should be included in the next release.

Full Changelog: v4.2.1...v4.3

v4.2.1

This update does not contain any new features.

The changes made were non-functional code cleanups.

Full Changelog: v4.2...v4.2.1

v4.2

This update contains an important bug fix, along with a variety of improvements and new features.

What's Changed

• Fixed a bug in AI Chat which caused lag or out-of-memory crashes in certain cases.
• Added support for the Browser Extension: The new "Ask About Browser Tab" command is disabled by default, and you can manually enable it in the Preferences. Also added the {browser-tab} dynamic placeholder.
• Added the Format Text command to intelligently fix the formatting of selected text.
• Added a cursor icon during response loading - just like previously used in ChatGPT. (cosmetic only, can be disabled in Preferences)
• Added the "Always Use Default Language" option in Translate.

Full Changelog: v4.1...v4.2

v4.1

This version brings huge improvement to Web Search, redesigned icons, a new model, and many bug fixes.

What's Changed

• A major rewrite of the Web Search feature: You can now enable Web Search for AI commands like Ask AI. Read more about it here!
• Redesigned many extension icons: The new logos are more aesthetic and consistent.
• Added the llama-3.1-nemotron-70b model.
• You can now set a default AI preset - simply edit a preset and check the box.
• Added support for more OpenAI-compatible APIs.
• Automatically Check for Updates is now enabled by default.
• Various cosmetic changes and bug fixes.

Full Changelog: v4.0...v4.1

v4.0

The next major version of raycast-g4f is here!

What's Changed

• New model
  • meta-llama-3.2-11b-vision
• New features
  • We now support file uploads for all providers, limited to text-based files. Providers like Google Gemini will still support all file types.
  • Added the Duplicate Chat action.
  • Allow editing uploaded files in Edit Message.
  • Stream responses in View Response.
• Many bug fixes
  • Fix system prompt not working.
  • Fix Default Language bug.
  • Small fix for Smart Chat Naming.
  • Fixed Google Gemini provider not accepting PDFs.
  • Fixed some chat settings not being saved.
• Cleanups
  • Removed Rename Chat action - it is superseded by Chat Settings.
  • Refactored some AI Chat code.
  • Deprecated some non-functional models.

Compared to v3.0, there have been numerous large improvements, including:

• Web Search is much faster and now accessible for all users.
• Added 5 providers and many new models.
• Added support for custom OpenAI-compatible APIs.
• Fixed automatic updates not working for some users.
• Many new features (check out the release notes for details)
• UI improvements.
• Many optimizations.

Lastly, this release also marks the 300th commit, as well as more than 300 stars, in this repository. A BIG thank you to everyone who used my extension, and I hope you find it useful! Your opinions help me prioritize features and fixes, so keep the feedback coming!

As AI progresses, more updates will come - stay tuned!

Full Changelog: v3.9.1...v4.0

v3.9.1

This is a bugfix update.

• Fixed automatic updates not working for some users.

What's Changed

• Fix automatic updates by @XInTheDark in #123

Full Changelog: v3.9...v3.9.1

v3.9

This update brings a huge improvement to the Web Search feature, as well as other enhancements.

Web Search has been made easily accessible for all users. We now use the DuckDuckGo search engine when the Web Search option is enabled. It's fast, private, and 100% free; therefore it has replaced our previously-used search provider. DuckDuckGo also doesn't need an API Key to access, so it's much more convenient.

Web Search is an incredibly useful feature, preventing hallucinations from the model, as well as providing the latest information. To try it out, simply check "Enable Web Search" in the extension preferences!

Other changes:

• Reverted the temporary patch in v3.8.1 - the issue has been fixed.
• We now allow editing (and resending) any message in the conversation, not just the latest one.

What's Changed

• Use DuckDuckGo for web search by @XInTheDark in #122

Full Changelog: v3.8.1...v3.9

v3.8.1

This is a bugfix update.

• Fixed a bug in the Blackbox provider where the order of messages would be reversed by the provider.
• Improved the system prompt for the "Summarize" command.

Full Changelog: v3.8...v3.8.1

v3.8

This version features many new additions!

• Added the ChatgptFree and Phind providers.
• Added the new meta-llama-3.2-90b-vision model.
• Updated Google Gemini models.
• Numerous improvements and bug fixes behind the scenes.

In past updates, many of these new features/providers were inspired by user-submitted feature requests. If you have an idea to improve the extension, please share it with me by opening an issue on GitHub!

What's Changed

• Add Phind provider by @XInTheDark in #119
• Add ChatgptFree provider (gpt-4o-mini) by @XInTheDark in #120
• Add meta-llama-3.2-90b-vision model (hosted by DeepInfra) by @XInTheDark in #121

Full Changelog: v3.7...v3.8

v3.7

This update brings support for a custom OpenAI-compatible API. You can also customize it with your own API Key and config. Getting started is simple - please visit this help page for instructions.

The default provider was changed to Nexra GPT-4o, as the previous default model is non-functional. If you're using an older version, and you're facing problems with responses, please try changing your default provider manually.

Lastly, as usual, more models and improvements have been added.

Author's note: Llama 3.2 just dropped less than a day ago - and likely many more to come! Looking forward to future advancements in open source AI, and of course, this extension.

What's Changed

• Support custom OpenAI-compatible endpoints by @XInTheDark in #118

Full Changelog: v3.6...v3.7