The scenario reads the site's pages via htmlExtract into a Supabase knowledge base, then answers visitors in chat from that content.