Add an optional LLM API call that tidies the output, either at the end, or for each page? happy to implement. Would be helpful.