I wrote an autonomous LLM-powered agent using LangGraph for HeyQuip which was capable of hunting down technical documents on the internet for rental equipment given a manufacturer and a model number. The agent uses a cognitive architecture that is capable of reasoning about the documents in order to assess accuracy, such as whether or not the data is from a primary source, or whether it is a useful or real technical document versus a marketing brochure, among other skills.
Once it has definitely found the document, it then extracts all relevant information from it, finds imagery related to the equipment, and then compiles it all into a JSON format including SEO data for use on rental websites.