Returns the AI metadata fields populated for the document. Each result combines the field definition (id, key, field_type, settings) with the extracted value (value, acceptance_status).

Only fields that have an extracted value for the document are returned.

The endpoint signals AI extraction state via the HTTP status code:

• 202 extraction_pending — extraction is in progress. Clients
  should retry after the number of seconds indicated in the
  Retry-After header.

• 204 No Content — extraction has terminally failed for this
  document. No body is returned; retrying will not help — contact
  support.

• 409 not_started — extraction has not been triggered for this
  document (typically because the document is not yet completed).