I am able to do it in python but I don’t understand why we can’t do it in the scrape source https://www.gumloop.com/pipeline?workbook_id=wj9uLit6jYjgPYwK8XaAU4. Ideally we should be able to traverse a dom elements without spending anything on the llm credits. I also think this might be a feature worth adding don’t run the LLM content on the whole page source specially when we have a lot of page with heavy html elements(wasting tokens). Let the end user define the scope of the html element and then run llm on top of those for extracting data