Best fit
- product listings
- company or job pages
- invoice-like HTML
- scraped web content that needs a typed output
Recommended stack
- model: Schematron
- output control: structured outputs
- scale path: batch
Workflow
- isolate the most relevant HTML if you can
- define the JSON schema you want back
- run Schematron against the HTML
- validate the response shape
- move to batch when the workload grows