{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/d341d94f3abd45628fc953c53da6cdfd\" frameborder=\"0\" width=\"1908\" height=\"1431\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1431,"width":1908,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1431,"thumbnail_width":1908,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/d341d94f3abd45628fc953c53da6cdfd-7187f234d1e7f793.gif","duration":151.566633,"title":"How a Shopping React Agent Validates Recommendations","description":"In this Loom I walk through a pipeline that turns a parent’s pre tick shopping question into grounded, milestone based product recommendations in English and Arabic. It uses three tools: milestone lookup, catalog search with scope checking, then JSON extraction and schema validation. I show real inputs like a My baby is 6 month old and we are starting slowly in the next week, and it returns catalog grounded options, for example a product with 95 percent confidence. Schema validation explicitly flags failures to the viewer. There is no action requested from viewers."}