gpt-4-vision-preview integration

[TimeLordRaps]

Is your feature request related to a problem? Please describe.
No

Describe the solution you'd like
Integrate gpt-4-vision and more generally visual language models using LangChain, by first grabbing a screenshot of the application, and then use https://github.com/reworkd/tarsier to tag the image, allowing the visual model to navigate the app.

Describe alternatives you've considered
Not using tarsier, but then the models aren't pseudo-grounded.

Additional context
I'm interested specifically in the CogAgent and SeeClick models currently as they seem to be performing best on visual langauge modeling tasks for web app navigation.