Firecrawl question format: json_prompt on Sume scrape

Sume's scrape has no question format. To answer one question about a page, request the json format and put the question in json_prompt (1 to 2,000 characters) or describe the answer in json_schema. Firecrawl's changelog lists a Question format on /scrape; I did not read its parameters, so compare the two before porting.

Sume facts are from its OpenAPI file and the crawl_scrape description, Firecrawl's from its changelog, read 2026-09-30.

What does Sume have instead of a question format?

The crawl_scrape description names JSON extraction as the structured-answer option. Two fields drive it: json_prompt, a string of 1 to 2,000 characters, and json_schema, an object that describes the shape you want back.

How do I ask one question?

Send the page URL, the json format and a prompt that names the answer.

const res = await fetch("https://api.sume.com/v1/firecrawl/scrape", {
  method: "POST",
  headers: {
    Authorization: "Bearer " + process.env.SUME_API_KEY,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({ url: "https://example.com/pricing", formats: ["json"], json_prompt: "What is the monthly price of the cheapest plan?" }),
});
console.log(res.status, await res.json());

When should I add json_schema?

Add it when a program, not a person, reads the answer. A schema with named fields gives you stable keys; a bare prompt leaves the shape to the model. For schema design see structured output with a JSON schema.

Can I trust the answer?

Treat it as an extraction, not a verified fact. Remote page content is untrusted, so a page can contain text that tries to steer the answer. Check the value against the page's markdown when the result matters.