MarkUDown

Create Schema API — MarkUDown AI extraction schema generator docs

POST/api/create-schema

Create Schema

Describe the data you want in plain language and get back a suggested extraction setup, generated by an LLM: the fields and their types, the extraction scope and keyword/path filters for URLs. The request is synchronous.

ParameterTypeRequiredDefaultDescription
querystringrequired—Plain-language description of the data you want, optionally with the site URL.

Request

curl -X POST https://api.scrapetechnology.com/api/create-schema \
  -H "X-API-KEY: your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"query": "title, price and star rating of every book on https://books.toscrape.com"}'

Response

{
  "url": "https://books.toscrape.com",
  "extraction_scope": "list_page",
  "extraction_target": null,
  "title": "string",
  "price": "float",
  "rating": "string",
  "allowed_words": ["catalogue", "book"],
  "blocked_words": ["login", "basket"],
  "allowed_patterns": ["/catalogue/"],
  "blocked_patterns": []
}

Python — pip install markudown

from markudown import MarkUDown

md = MarkUDown(api_key="your-api-key")
schema = md.create_schema("title, price and star rating of every book on https://books.toscrape.com")
print(schema)

Node.js — npm install markudown-sdk

import { MarkUDown } from "markudown-sdk";

const md = new MarkUDown({ apiKey: "your-api-key" });
const schema = await md.createSchema("title, price and star rating of every book on https://books.toscrape.com");
console.log(schema);
The response is one flat JSON object: every key other than url, extraction_scope, extraction_target and the four filter lists is a field name with its type. To use it with Extract, turn each field into {"name": …, "type": …, "active": true}. extraction_scope is one of whole_site, category, single_page, search_query, list_page or unknown. Review the output: it is written by an LLM.

Cost: No credits charged; counts toward the rate limit.