Skip to main content

​
Short answer

Billing data only

Logs contain timestamps, model, token counts and amounts — the fields billing needs. Your inputs and outputs are never stored.

Rolling last two months

Cleanup runs on the 5th, 15th and 25th of each month, each removing data older than two months. Example: 15 August clears everything before 15 June.

Export for long-term records

Use async export in the Logs section of the console and archive the file yourself for reconciliation and audits.
Cleanup is irreversible. Cleared logs cannot be recovered. If you need them for reconciliation, audits or expense claims, export before the cleanup date.

​
What the logs contain

We keep only what billing and troubleshooting require. That is both a privacy-minimization principle and a direct reduction of the surface area for any data leak. For what each column means, see how to read the billing amounts in the logs.

​
Can I see generated images and videos in the console?

This is the most common follow-up question, and the answer depends on how the call was made:

Synchronous calls: no

Results from synchronous text and image endpoints are returned straight to your program and never stored. The console sees neither the prompt nor the generated image.

Async tasks: yes, but links expire quickly

Video and other async tasks work on a submit-then-poll model, so the result must be held temporarily to be queryable — which means the artifact is visible.
OSS links for async task artifacts last about one day. Video and image download URLs expire after roughly 24 hours and cannot be retrieved afterwards. Transfer results to your own storage as soon as you receive them — do not use these temporary links in production or hand them to end users.

​
The cleanup schedule

Logs are kept on a rolling window, on a simple rule: cleanup runs on the 5th, 15th and 25th of each month, and each run deletes data older than two months — three runs a month, leaving roughly the last two months queryable at any given moment.
Because a run happens every ten days and cuts by date rather than by whole month, the usable window drifts between two months and two months plus ten days — shortest right after a run, longest just before the next one.
This page is the notice; individual runs are not announced. Cleanup happens automatically on the fixed schedule above, with no separate announcement before each run — treat the rules and table on this page as the reference.
Log tables grow fast, and deleting a whole month in one pass is large enough to slow the database down and affect live calls. Running a smaller batch every ten days keeps both the operational load and the impact on live traffic low. That is why the schedule changed from one whole-month cleanup per month to three rolling cleanups per month, effective 15 August 2026.
Cleanup only removes call-log detail; your account balance is unaffected. If you also need long-term records of top-ups and reconciliation, export and archive those as well.

​
Keeping logs longer

1

Open the Logs page

Go to https://api.apiyi.com/log, or click Logs in the top navigation.
2

Filter first

Narrow by time range, model or status so you export only the records you actually want to archive.
3

Use async export

Click async export to start the job. It runs in the background, so large exports do not require you to sit on the page.
4

Download and archive

Download the file once the job finishes and store it in your own system or finance workflow.
Make monthly exports a habit: archive the previous month at the start of each month and the cleanup date stops mattering. This is especially worth doing for teams with reconciliation, expense-claim or audit requirements.

​
Common questions

Two reasons: privacy minimization — the less retained, the smaller the exposure; and storage cost — log volume grows fast on a high-traffic platform, and keeping everything forever eventually shows up in pricing. We keep a window sufficient for reconciliation and give you the export tools to retain more yourself.
No. Cleanup is a physical delete with no recovery path, and support cannot restore it either. Export before the cleanup date.
No. Cleanup targets call-log detail; your account balance is unaffected. If you need long-term proof of top-ups and spending for expense claims or audits, export and keep those too.
Within the retention window, the timestamp and model name from the log are enough to locate the vast majority of issues. If you genuinely need more detail, contact support and an administrator can temporarily enable detailed logging — see can I view detailed logs in the backend.
No. The export contains the log itself, with the same fields shown in the console — time, model, token counts, amount, status — and no prompt or model output whatsoever.
Enterprise customers with specific retention or compliance requirements can discuss it with our sales team. The standard recommendation is still regular exports — see enterprise vs individual users.