Comment by klodolph

I don’t understand why the agents requesting HTML can’t extract text from HTML themselves. You don’t have to feed the entire HTML document to your LLM. If that’s wasteful, why not have a little bit of glue that does some conversion?