v2.1

Added 3 new models: Mixtral-8x7B, Gemma-2-27B and OpenChat-3.6-8B.

What's Changed

• Add Mixtral-8x7B and Gemma-2-27B models (hosted by DeepInfra) by @XInTheDark in #65
• Add OpenChat-3.6-8B model (hosted by DeepInfra) by @XInTheDark in #66

Full Changelog: v2.0...v2.1

v2.0

Raycast-g4f v2.0 is here!

We introduce the following changes:

• Added new model Mistral-7B.
• Web search functionality for DeepInfra models have been completely rewritten and updated to support native function calling. This allows for a more seamless integration and should provide much better user experience.
• New feature Code Interpreter: Allows GPT to execute Python code locally, to handle complex tasks (e.g. math, data analysis). The model has been instructed to strictly only produce safe code, but use at your own risk. Currently, only selected DeepInfra models support this feature.
• Major changes to the codebase, which should make the extension faster and more lightweight. For example, we have reduced the dependencies used.
• Numerous bug fixes and QoL improvements.

⚠️ IMPORTANT NOTE: This update introduces major changes to data formats. Some data from previous versions of the extension are now incompatible. If you update to this version, your AI Chat data will be lost, unless you back them up using the "Export Chat" function. If you experience issues launching commands after updating, it's likely because of the data incompatibility. Please follow the troubleshooting instructions here.

What's Changed

• Remove Qodana CI by @XInTheDark in #60
• Improve web search for DeepInfra models, restructure message data by @XInTheDark in #59
• Add Mistral 7B model (hosted by DeepInfra) by @XInTheDark in #61
• Add Code Interpreter feature by @XInTheDark in #62
• First step to removing g4f-ts package by @XInTheDark in #63
• Remove g4f-ts package, replace with g4f-image by @XInTheDark in #64

Full Changelog: v1.5...v2.0

v1.5

In this release:

• The new Qwen2-72B state-of-the-art model, hosted by DeepInfra, has been added!
• The now-inactive Dolphin 8x7B model has been retired.
• The "Provider" option has been added to the G4F Local API config.
• Bug fixes for file uploads.
• QoL updates.

What's Changed

• Bug fixes for file uploads by @XInTheDark in #57
• Add Qwen2-72B model (hosted by DeepInfra), remove Dolphin 8x7B by @XInTheDark in #58

Full Changelog: v1.4.1...v1.5

v1.4.1

A new preference was added: Persistent Storage. It enables more persistent storage of the extension's data, like AI Chat data or Custom Commands, by backing up a copy of the data to storage files on your computer. It's experimental and disabled by default, but has undergone a very large amount of testing.

Also includes a few cosmetic changes.

What's Changed

• Add Persistent Storage preference by @XInTheDark in #56

Full Changelog: v1.4...v1.4.1

v1.4

The next major update is here!

To start off, introducing the next huge leap for the project: The extension now supports integration with the locally hosted API of the gpt4free project. You can easily host your own gpt4free API locally, and connect it to the extension. This opens up many possibilities, allowing for a great variety of up-to-date providers. These include Meta AI, Bing Chat and more.

Setup is simple and the steps can be found in the dedicated help page.

There were also large improvements made to Dynamic Placeholders, and the newly-added documentation can be found here.

Lastly, as promised, the Privacy Policy was added. I want to be completely transparent about how the extension uses your data, and this document explains it clearly.

Thank you for using raycast-g4f, and more features are coming soon!

What's Changed

• Create Privacy Policy by @XInTheDark in #52
• Add more Dynamic Placeholders by @XInTheDark in #53
• Add support for GPT4Free Local API by @XInTheDark in #54

Full Changelog: v1.3...v1.4

v1.3

This is an important bugfix update and is strongly recommended for all users.

This update fixes the following issue:

When running AI commands, there would often be two responses being generated in parallel. This would lead to the displayed text switching back and forth between two different versions of the response. While this did not affect the final result, the user experience could be degraded.

This issue has always been present in the extension, as well as in the original codebase.

What's Changed

• Fix double response generation in AI commands by @XInTheDark in #51

More features soon

On a separate note, I'm currently working on quite a few enhancements that will bring even more advanced functionalities and integrations. If you'd like to try out these updates when they're available, consider enabling "Automatically Check for Updates" in the preferences!

Full Changelog: v1.2.2...v1.3

v1.2.2

In this update:

• The "Generate Images" UI has been reimagined and improved, and it should now be more familiar to use!
• Additionally, you can now choose from over 50 different image generation models.
• QoL updates, including: A keyboard shortcut was added for "Pin Chat".

What's Changed

• Revamp "Generate Images" by @XInTheDark in #48
• Allow model selection in "Generate Images" by @XInTheDark in #49

Full Changelog: v1.2.1...v1.2.2

v1.2.1

You can now pin chats in AI Chat by selecting "Pin Chat" in the actions! This will:

• Display the chat in a separate "Pinned" section in the chats dropdown;
• Prevent the chat from being purged If the auto-delete chats feature is enabled.

To unpin a pinned chat, just select "Pin Chat" again.

What's Changed

• Pin Chat functionality in AI Chat by @XInTheDark in #47

Full Changelog: v1.2...v1.2.1

v1.2

The next version of raycast-g4f is here! Major changes include added support for uploading files: this includes AI Chat, several AI commands, as well as optionally in Custom AI Commands. To upload a file in AI Chat, press Command-Enter or select "Compose Message" from the actions.

The new Ask About Screen Content command is also introduced - it instantly takes a screenshot and sends it to the AI along with your prompt.

Currently, both features are limited to the Google Gemini provider only, but I expect to add more providers in the future as the trend shifts towards multimodality. Parts of the code base have already been rewritten in order to pave the way.

Do give it a try! Accessing Google Gemini is easy and completely free, and the steps are found in the README.

Lastly, under the hood, there have been numerous code updates and QoL improvements.

What's Changed

• Add Storage interface for persistent storage by @XInTheDark in #43
• Rewrite code: Introduce Message & MessagePair class by @XInTheDark in #42
• Switch to gemini-g4f package by @XInTheDark in #36
• Add support for uploading files by @XInTheDark in #44
• Add "Ask About Screen Content" command by @XInTheDark in #45

Full Changelog: v1.1.1...v1.2

v1.1.1

More improvements to the auto-update feature! Now, there is another (experimental) feature that allows the extension to periodically check for updates every day. To try it out, simply check the box in the Preferences for "Automatically Check for Updates".

What's Changed

• Automatically Check for Updates by @XInTheDark in #41

Full Changelog: v1.1...v1.1.1