
When Imagild returns HTTP 429, stop immediate retries, respect Retry-After, and reschedule the same intended operation. The current gateway returns Retry-After: 60. A throttled request does not prove that a previously accepted generation failed, and changing its idempotency key can create an unintended additional generation.
Key Takeaways
- Current account budgets are 10 generation/upload writes and 120 reads per minute.
- Cool down the affected request class instead of letting every worker retry independently.
- Preserve the original operation key and request body when retrying a generation.
Identify which request was throttled
Record the method, endpoint, response status, error code and error.request_id. Keep API secrets out of logs. A 429 from GET /tasks/{id} concerns that status request; it is not the generation task's final state. A 429 from a new submission is different from a timeout after a submission whose outcome is unclear.
The Imagild API console is the product reference for current limits and task handling. MDN's 429 reference explains the general meaning of throttling and why its scope depends on the service. Do not assume that an error number reveals a provider's private queue or remaining capacity.
Convert the retry hint into a deadline
Imagild currently supplies a delay in seconds. This small application helper handles that current form and uses a conservative fallback when the header is absent or malformed:
function retryDeadline(retryAfter, now = Date.now()) {
const raw = retryAfter?.trim() ?? "";
const seconds = /^\d+$/.test(raw) ? Number(raw) : 60;
const safeSeconds = Number.isSafeInteger(seconds) ? seconds : 60;
return now + Math.max(60, safeSeconds) * 1000;
}
const notBefore = retryDeadline(response.headers.get("Retry-After"));
// Persist notBefore with the operation; do not send again before it.
This helper deliberately handles numeric delays, not every possible HTTP-date form. MDN documents both Retry-After formats. Add date parsing if your general-purpose client needs it. A retry deadline is permission to try later, not a guarantee of immediate acceptance.
Pause the scheduler instead of multiplying timers
Use a shared cooldown in your own account scheduler. When one write is throttled, stop dispatching other pending writes until the cooldown and your local rate budget allow them. Reads have a separate budget, but they still need coordinated pacing.
Add a small randomized delay after the minimum wait so restarted workers do not all resume in the same instant. Store the next eligible time durably. A browser refresh or worker restart should not erase the cooldown and release the whole backlog.
These are application controls. They are not an Imagild distributed-lock interface, a queue-position endpoint or a promise that every server instance maintains one globally exact quota counter.
Retry an operation rather than recreating the work
For a generation retry, keep the existing Idempotency-Key, payload and known task ID. If a task ID already exists, inspect that task rather than creating a replacement just because a status read was throttled. If the task reports unknown, preserve its identity and seek reconciliation.
Do not treat a confirmed validation error or expired membership as something a longer sleep fixes. Inspect the actual error and correct the relevant condition. The failed-job next-action guide helps separate targeted repair from blind resubmission.
Verify recovery with controlled evidence
Test the cooldown locally with mocked responses before connecting paid generation. Confirm that several waiting workers produce one coordinated retry schedule, no request starts before its deadline, and operation keys remain unchanged. Include a second 429 to ensure the client can extend its wait.
In production, measure your own throttled requests and successful recoveries without calling them a service-wide availability benchmark. Active membership and sufficient points still apply after throttling clears. Use the points and jobs guide for the spending context. Pass the recovery flow when the intended task resumes without a burst, duplicated submission or falsely reported task 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: Current request limits, retry handling, authentication and asynchronous task lifecycle.
- MDN HTTP 429: General throttling semantics and service-specific scope; no Imagild provider-capacity claim.
- MDN Retry-After header: Numeric delays and HTTP-date forms; the example explicitly handles the current numeric form.
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.