Skip to main content
To see how your visibility changes, run the same prompts on the same models every day. Store each result, then compare the days. This recipe runs 3 prompts on 3 models: 9 tasks a day, $0.117 a day, about $3.51 for 30 days.

1. Create the tasks every day

daily.mjs
  • The Idempotency-Key names the day, the prompt and the model. If the job runs twice, or you rerun it after an error, no task is charged twice.
  • metadata comes back on the task and in the webhook. Use it to file each result under its prompt set and day.

2. Store each result

The webhookUrl gets task.completed for each task. Verify the signature, then save what you need: the result is readable for 7 days only.
verifyWebhook is in Verify webhook signatures.

3. Watch for failures

A failed task is refunded, and its webhook is task.failed. To find the failures of a day, list them:
To try one again, create a new task with a new Idempotency-Key, for example with -retry1 at the end. Next: Compare brands across models.