BrowserOS Agent - v50.0.0.3

[shadowfax92]

✨ What's new:

• Agent v3 is here! – Our browsing agent just got a major upgrade to v3, now much smarter at handling complex tasks and workflows.
• Enhanced local model support – Now using combined Visual + semantic DOM for better performance with local models.
• UI improvements – Various interface enhancements. Thanks to @OmkarBansod02 for the contributions!

---

This discussion was created from the release BrowserOS Agent - v50.0.0.3.

[MentalGear]

Congrats on the release!

I been wondering how you sandbox the AI in the browser, a quick look at the readme didn't surface anything definite.
E.g. something like DeepMind's CAMEL (Dual LLM approach?).
But you should be aware that there is a new sandbox approach you might find useful:
https://www.earlence.com/blog.html#/post/cellmate