Open
Description
we could possibly take a screen shot of large HTML outputs, especially when they are interactive with javascript code, instead of relying on the raw content of the output.
This would likely reduce the token count required, and also be a more accurate representation of the output in its current state