🎉 Release notes for the Windows version (v7.1):

v7.1 Hotfix Additions:

• Fixed an issue with the OpenAI Provider, and improved & optimised the entire AI Provider architecture.
• Improved the thinking text animation.
• Fixed the default icon arrangement to be the same as v6.

v7 Additions:

1. Customizable buttons! You can now add, edit, remove, and rearrange the buttons.

Huge thanks to @momokrono for building the entire back-end for this 💛! I added the GUI (with draggable buttons, reset to default functionality...).

2. Sleeker pop-up UI

3. A native Ollama AI Provider (using the Ollama API), for easier Ollama setup & custom model-in-memory-time options — entirely thanks to @momokrono!

4. New Gemini model options, with better explanations

5. A new Pause & Resume option to stop Writing Tools from reading your hotkey if necessary. Again, @momokrono so awesomely implemented this :D

5. Other Improvements:

• UX improvements to the Summary etc. pop-up window (optimized initial height)
• Friendlier error message on hitting Gemini rate-limits
• Added graceful quitting when running through the terminal [credit: @momokrono]

I want to thank @momokrono once again for all his help and support—not only in contributing code, but also for extensively helping manage GitHub issues. Thank you, truly—I'll never forget your kind contributions. 💛

Upcoming features:

• Localization (multi-language UI) — @momokrono has built the entire framework for this, and my gratitude is unbounded — this would help so many. It'll be ready by the next release!
• Hotkeys for specific options.
• Support to remove reasoning token markers if needed (PS: DeepSeek R1 & Gemini 2.0 Thinking already work!).
• (researching) Tracked changes/diff check (to compare proofread text with the original).
• Checking that the active window is the same before pasting.
• More feature suggestions on GitHub Issues, and potentially future ones.
• And of course, bug fixes and improvements overall :]

⬆️ Update Instructions:

You can keep using your old config.json to preserve your set-up! Simply bring it over into the new Writing Tools v7.1 folder you'll download below.

🐞 Known Issues

1. On some devices, Writing Tools may not work correctly with the default hotkey.
   To fix it, simply change the hotkey to ctrl+` or ctrl+j and restart Writing Tools.
   PS: If a hotkey is already in use by a program or background process, Writing Tools may not be able to intercept it. The above hotkeys are usually unused.

2. The initial launch of the Writing Tools.exe might take unusually long — this seems to be because AV software extensively scans this new executable before letting it run. Once it launches into the background in RAM, it works instantly as usual.

🆕🎉 Release Notes for the macOS Version by @Aryamirsepasi (v4.1):

• Revamped Local LLM page: You can now select either an LLM with OCR or a VLM with direct image recognition.
• New Local AI Models: Qwen 3.0 4B, Gemma 2 2B, Qwen2.5 VL
• Fixed Permission Issue: Fixed the annoying bug that plagued previous versions. For future updates you no longer have to reset the permissions after each update. However, if you are using a previous version, you would still have to re-enable accessibility access one last time:

System Settings → Privacy & Security → Accessibility → Select writing-tools → Press the minus (–) button → Press the plus (+) button → Add writing-tools again and enable access.

v4 Release Notes:

• Integrated AIProxySwift for all providers.
• Image Recognition in Finder: Simply open WritingTools on the desired image in Finder, and it will be recognized by the app.
• Image Support for All Providers:
  • Some providers like Google and OpenAI have native image support via their APIs, which is fully utilized here.
  • Others, like Mistral and Local LLM, use OCR and Apple Vision for text recognition on images.
  • For Ollama, you can choose between native support (if your AI model supports it) or OCR via Settings.
• Keyboard Shortcuts for Each Tool:
  You can set up separate shortcuts for each tool in its Edit Page.
  (Note: Shortcuts will only work if the app is running in the background.)
• Font Size Options have been added back to the Response Window.
• New Gemini models, plus a custom option to add your own Google model name for future use.
• UI Improvements & Bug Fixes.
• Spanish Translation added.
  (If you encounter any translation issues in any language, please let me know.)

Please Note:
After updating the macOS version, the tools may sometimes not show up.
To fix this, re-enable accessibility access:

System Settings → Privacy & Security → Accessibility → Select writing-tools → Press the minus (–) button → Press the plus (+) button → Add writing-tools again and enable access.

Upcoming Features:

• Text streaming in the response window.
• Better support for reasoning models.
• Gemma 3 support as soon as it is completed by the MLX team.
• Support for videos and links via Gemini.
• Still testing: evaluating the possibility of using a Core ML or llama.cpp model instead of MLX — which should, theoretically, allow Local LLM to run on Intel Macs as well.