
A polling deadline tells you that your application stopped waiting; it does not establish that the Imagild generation failed. Keep the existing task ID and read it again when the connection or wait window allows. Change the displayed waiting state without inventing a failed provider outcome.
This is particularly useful for users who leave a page open on an unreliable connection. Their screen can lose contact while the accepted task still has a separate server-side lifecycle. Start recovery from the Imagild API console's integration contract, not from a new generation button.
Key Takeaways
- Store a local wait outcome separately from
data.status. - Resume the existing task before considering any new paid operation.
- An unreceived submission response needs the original operation identity, not a fresh idempotency key.
Identify which clock expired
The word “timeout” can refer to several events. Record which one occurred before choosing a recovery action.
| Event | What your application knows | First action |
|---|---|---|
| A two-minute UI wait budget expires | The screen has stopped waiting | Retain task ID and offer another status check |
| One task-read request times out | That read did not produce a usable response | Retry a later GET within a bounded policy |
| A submission response is lost | Acceptance may have occurred | Recover with the same operation key and unchanged parameters |
| Task JSON reports failed | A generation failure is now reported | Inspect error and usage state |
| Task JSON reports unknown | The upstream outcome is unresolved | Preserve the record for reconciliation |
A provider deadline can mean something different again. Replicate documents provider-side prediction deadlines; those controls and billing outcomes do not establish a corresponding Imagild cancellation feature.
Write a recovery record that does not overwrite evidence
Suppose your own application waits two minutes for a 2K product image. Save separate fields such as these in your database:
{
"operationId": "product-banner-20261028-0042",
"taskId": "11111111-1111-4111-8111-111111111111",
"lastRemoteStatus": "processing",
"lastObservedAt": "2026-10-28T09:00:00Z",
"localWaitState": "deadline-exceeded",
"nextAction": "read-existing-task"
}
The UUID and timestamp are illustrative. Your application owns the surrounding field names; they are not extra fields to send in the generation request. A local timeout should update localWaitState, not turn lastRemoteStatus into failed.
When the user returns, your server reads GET /tasks/{taskId} with its Bearer key. Valid membership is still required. If that read now returns succeeded, proceed to content download. If it remains processing, show the latest observation time and allow bounded waiting again.
Recover a lost acceptance response differently
If you never received data.id, a GET cannot use an ID you do not have. Locate the operation record created before submission. Reuse its original Idempotency-Key with the same parsed generation parameters at the original endpoint.
Imagild reuses the matching task rather than creating a separate operation. Changed parameters under that key return a conflict. Do not append a new timestamp or switch models as an automatic timeout workaround; those changes would describe a new generation and need deliberate budgeting.
This distinction separates recovery from creative revision. The points and job guide can help explain why a revised output is a separate spending decision.
Verify timeout handling without buying duplicate results
Use an application-level fake reader that returns processing after the UI deadline, then success on a later visit. Confirm that the same task card resumes, its original model and quote remain attached, and the submission counter stays at one.
Also simulate a 401 or membership-related 403 during recovery. These are access problems; neither proves the media task failed. Leave the remote outcome unchanged while asking the user to restore access through the appropriate account process.
For unknown, keep the task and sanitized diagnostics for review. For confirmed failure, follow the failed-job next-action guide. The practical result is a truthful interface: “waiting stopped” remains recoverable, while a failure label is reserved for an actual reported generation failure.
Related tasks
- What Should You Check Before Retrying a Failed Image or Video Job?
- Imagild Points, Resolution and Jobs: Plan a Creation Before You Spend
Sources and scope
- Imagild API console and integration guide: Current Imagild entry point, active-member access, asynchronous task reading and authenticated output contract; implementation facts cross-checked in lib/relay and the 2026-10-28 claims manifest.
- Replicate Prediction Lifecycle: General distinction between an ongoing prediction and its reported outcome. Its provider deadlines, cancellation, billing and retention rules are not attributed to Imagild.
Product instructions were checked against the current Imagild API contract and the public model catalog. Code, budgets and acceptance checks are illustrative; this article does not claim a new paid generation test, an uptime measurement, a search-volume study or a ranking of AI citations.
Prepared by Imagild Editorial, the company editorial team responsible for these guides. AI assisted research and drafting. This is documentation-based guidance with proposed examples, not a claim of hands-on testing or an independent product review. Report a correction.
Cover: Original Imagild product screenshot captured on 28 October 2026; model availability and prices may change. This image is not an API response or a performance benchmark.