Retry without duplicating work
Recover the original request before deciding to start another paid extraction. A lost connection is not evidence that the service rejected the request.
Preserve the original identity
Keep the operation ID, URL, options, original request body and original Idempotency-Key. Do not put credentials in your recovery notes. The worksheet below is a personal record, not an API request or service configuration. Its uppercase entries identify values to record, not literal service values.
service: SERVICE_NAME
operation_id: OPERATION_ID
idempotency_key: ORIGINAL_KEY
request_body_file: path/to/original-request.json
version: EXACT_VERSION_IF_APPLICABLE
last_observed_state: STATE
Recover in the workbench
- Keep the current Page request open. Do not clear it with New request while acceptance is uncertain.
- If an operation ID was received, choose Check existing operation or Check selected request. Confirm the same ID and URL.
- If only the submission was retained, choose Check existing submission. This checks the original body and request identity rather than starting an unrelated request.
- If a completed result is returned, inspect it. If the operation failed, read its saved error before deliberately submitting new work.
API retry rules
Read an existing ID first. If no ID was received, replay the original body using the original key. A matching replay returns the stored operation; it does not guarantee success or rerun a failed extraction. A changed body with the old key can return idempotency_conflict. Restore the original body to recover that attempt; use a new key only when the URL or options intentionally define new work.
When to stop retrying
Authentication, paid entitlement, robots restrictions and private-URL refusals need their own correction. Repeating a request does not grant access. On service failure, retain the evidence and retry the read after recovery. Check the operation state reference before using any result.