Skip to main content
function extractArrayFromPage(page, options): Promise<Record<string, string>[]>
Extracts an array of structured data from a web page in an optimized way, this function will use ai for the first n times, until it collects multiple examples then it will build reliable selectors in the background to make the process more efficient

​
Examples

 import { extractArrayFromPage } from "@intuned/sdk/optimized-extractors";

 await page.goto("https://books.toscrape.com/")
 const books = await extractArrayFromPage(page,
   {
     strategy: {
       model: "gpt4-turbo",
       type: "HTML"
     },
     itemEntityName: "book",
     label: "books-extraction",
     itemEntitySchema: {
       type: "object",
       required: ["name"],
       properties: {
         name: {
           type: "string",
           description: "book name",
           primary: true
         }
       }
     }
   },
 )

 console.log(books)

 // output:
 // [
 // ...
 // { name: 'Olio' },
 // { name: 'Mesaerion: The Best Science Fiction Stories 1800-1849' },
 // { name: 'Libertarianism for Beginners' },
 // { name: "It's Only the Himalayas" }
 // ...
 // ]

​
Parameters

• page: Page The Playwright Page object from which to extract the data. • options • options.itemEntityName: string The name of the entity items being extracted, it must be between 1 and 50 characters long and can only contain letters, digits, periods, underscores, and hyphens. • options.itemEntitySchema: SimpleArrayItemSchema The schema of the entity items being extracted. • options.label: string A label for this extraction process, used for billing and monitoring. • options.optionalPropertiesInvalidator? Optional. A function to invalidate optional properties. • options.prompt?: string Optional. A prompt to guide the extraction process. • options.strategy?: ImageStrategy | HtmlStrategy Optional. The strategy to use for extraction, if not provided, the html strategy with claude haiku will be used. • options.variantKey?: string Optional. A variant key for the extraction process, use this when the page has multiple variants/shapes.

​
Returns

Promise<Record<string, string>[]> A promise that resolves to a list of extracted data.