How does huggingchat prompt the model to generate HTML output?

[vgoklani]

How does Huggingchat prompt the LLM to generate HTML output? Where can I find that prompt? I'd like to tweak it. thanks!

[nsarrazin]

We don't prompt it specifically to generate HTML. Do you have an example of what you mean specifically ?

[vgoklani]

Apologies, the question should be asking how are you formatting the text into HTML as its streaming. Thanks!

[nsarrazin]

Aah no worries. We use marked to parse the text as markdown, which we then render to HTML.

You can see the component that handles this logic here

Let me know if that helps :grin:

This is the block that renders it, and the marked tokens are generated here

[julien-c]

(hi @vgoklani! 👋 )