Skip to content

Create a Scraper

Scrapers are the right tool when you want item-level metadata collection with review before apply.

Build a scraper when the workflow is:

  • start from an existing item in the library
  • fetch metadata from a source
  • review the result
  • apply only the fields the user wants

If the workflow is primarily “bring new content into the library,” you probably want a downloader instead.

A good scraper should:

  • make source assumptions clear
  • fail visibly when the source structure changes
  • return data that is easy for a user to review
  • preserve enough source detail that provenance remains useful
  1. Choose the source and define the item types it supports well.
  2. Decide which fields the scraper should populate confidently.
  3. Run it against a small representative set first.
  4. Review the output inside Cove before broadening its use.
  5. Package it so Cove can discover and load it through the extension model.

Use the local dev loop with a small demo library and representative URLs or source items. Focus on:

  • field accuracy
  • missing-field behavior
  • partial matches
  • broken-source handling
  • user review quality inside the apply flow

Do not try to make a scraper silently do everything. Scrapers are strongest when they are reliable about a narrower source pattern and leave room for user review where ambiguity is normal.