For pages with no reliable selectors
Describe the fields in a schema and get validated JSON back
Send a page, a sentence about what you want and a JSON Schema, and get back JSON that has already been checked against it.
The problem
Event listings from a hundred venues, product pages from a dozen shops, press releases from every company in a sector. Each site marks its data up differently and changes it without notice, so a selector per site is a maintenance job that never ends.
How domscout handles it
POST /extract takes a url, or the text or HTML you already have, plus a prompt and a JSON Schema. A language model reads the page text and must answer in JSON, and domscout validates that answer against your schema before returning it, so a field of the wrong type or a missing required field never arrives as data. If the first answer misses the schema, the validation error goes back to the model once; a second miss is an unbilled 502. A url is fetched as plain HTTP without running JavaScript, so for a page that builds its content in the browser, capture it as Markdown first and send that as content.
Where it shows up
- Event, job or listing records from sites that all mark them up differently.
- Product details from many shops normalised into one shape.
- Structured fields from text you already hold: emails, notes, exported pages.
The request
POST /extract. /extract starts at Pro and costs 4 credits a request. Fetching the page and the model's answer share an 18-second limit, so sending a bounded excerpt as content leaves more of it for the model. The schema supports a stated subset of JSON Schema, and a keyword outside it is refused with a 400 rather than ignored.
Go further
More in Data extraction and search: Crawling a whole site · Structured data from web pages · Semantic search over your content · Content behind clicks · Company research from websites · Evaluation sets from real pages · All use cases
Give your product a browser.
Get clean web content and visual proof into your workflow in minutes.