Here are the major changes we've released in July 2025:
We've refined the action with a few new features:
You can now provide a predefined data_schema_id to parse your webpage content into or you can provide a data_schema JSON object directly to the action so you can parse content with a single request.
If you are only interested in a certain part of the page, you can now define a selector and only the contents of that will be parsed using AI.
You can now provide output_type which tells Gaffa to return the required data as a file or embedded in the request.
We've added the action which is really useful if you are trying to scrape a site that implements infinite scrolling where items are removed from the page when out of view, usually modern javascript web apps.
We've now added an llms.txt file for the docs so you can import all the Gaffa docs into your favourite large language model and get it to write your automations for you!