enable text-based browser action for models that don't support images

[edwin-truthsearch-io]

We should have a browser tool that will allow a powerful model like Grok 3 to be able to scrape data and even interact with web pages on the internet without having to use the UI. This can be done by accessing the DOM of a headless browser. This will unshackle models that have tools support but not images support (very many of them) to do internet research to come up with even better solutions.