Well that's what I implemented. There are markdown docs for every HTML file and the proxy decides to serve either markdown or HTML based on the Accept header.
There is no real reason to pass HTML with tags and all to the LLM - you can just strip the tags beforehand.