What I learned using local vision-language models to scrape target.com

Using a 2 billion local visual language llm coupled with playwright to identify and click on elements in a browser.