Create a Vimium fork to support multimodal models

[ishan0102]

It might make sense to create a fork of Vimium designed specifically for making it easier for multimodal LLMs to choose relevant elements on a page. This might involve messing around with annotation colors, sizes, fonts, etc.

[philc]

Vimium author here. I have no opinion about whether to fork. I just heard about this project today and wanted to say, this is cool! Good luck!

[ishan0102]

@philc Wow thank you so much, means a lot coming from you! I love your work!

[asim-shrestha]

We just open sourced a utility library that can tagify web pages for you: https://github.com/reworkd/tarsier

Could be a drop in replacement for vimium. We have plans to be able to customize tag appearance / positioning if that's interesting

[aincube]

Just an idea, but maybe possible to use qutebrowser :

• written in Python (using QtWebEngine)
• has vim-mode par default
• has userscripts - really neat feature to automate things