Actions15
LLM Extractor
AI-generatedSummary
Extract structured data from a specified webpage using instructions and a user-defined schema interpreted by a large language model (LLM).
Inputs
- URL (required) — The URL of the webpage to extract content from.
- Extraction Instructions (required) — Text instructions guiding the LLM on what specific data to extract from the page.
- Schema Input Mode (required) — Select between defining the extraction schema using simple fields or an advanced JSON schema.
- Schema Fields (required) — When using simple mode, define each field's name, type, optional description, and whether it is required.
- JSON Schema (required) — When using advanced mode, provide a JSON schema defining the expected extracted data structure.
- LLM Options — Options to configure the LLM input format, maximum tokens allowed in the response, model selection, and response creativity (temperature).
- Options — Additional extraction options including array handling strategies, whether to clean extracted text (normalize whitespace), and whether to include metadata or the original page text in the output.
- Browser & Session — Settings controlling the browser emulation used to load the webpage, such as user agent, browser type, headless mode, JavaScript execution, proxy configuration, cookies, viewport size, and anti-bot evasions.
- Crawl Settings — Parameters controlling the crawling behavior such as respecting robots.txt, caching mode, CSS selector to limit extraction scope, JavaScript to execute on page, delays, retries, and other advanced options.
Output shape
a single item with JSON data extracted according to the specified schema or fields
Returns structured extracted data corresponding to the user-defined schema fields or JSON schema. May include metadata and optionally original page content if selected. Supports configurable array splitting strategies in the extracted data. Operation relies on LLM-powered content extraction from a rendered page loaded with user-specified browser and crawling settings.
Examples
Example 1: Extract title and description from a product page
Provide the product page URL, instructions to extract product name and description, select simple schema mode, define schema fields for 'title' and 'description' as strings, and optionally configure LLM and browser settings.
Example 2: Extract complex nested data using JSON Schema
Provide the page URL, detailed extraction instructions, choose advanced schema mode, and supply a JSON schema outlining the nested structure expected in the extracted information, such as arrays of objects with typed fields.