Can resubmitting a timed-out image start create two tasks?
Resubmitting a timed-out POST /v1/images/tasks request does not create a second task when the retry reuses the original Idempotency-Key, request path, and exact JSON body within 24 hours. The gateway replays the stored 201 response with the Idempotent-Replayed: true header rather than queuing another render.
Safe recovery steps after a timeout
- Check local storage for a saved
task_id(data.id). When atask_idalready exists, queryGET /v1/tasks/{task_id}instead of callingPOST /v1/images/tasksagain. - When no
task_idreached the client and the original request included anIdempotency-Key(at most 255 characters), resend the identical path, header, and JSON body within 24 hours. - Persist the
Idempotency-Key, request body, and returnedtask_idtogether on the order or asset record so later retries stay traceable.
Conflict and new-key boundaries
- In-flight or mismatched requests (
409 idempotency_conflict): Reusing a key while the first request is still running, or pairing the key with a changed body or path, returns409 idempotency_conflict. Wait for the original request to finish before retrying an identical payload. - Failed validation or balance checks: BeatAPI stores idempotency records only for
2xxresponses. After a400or402refusal, fix the request or balance and submit again. - Distinct tasks: Generate a fresh
Idempotency-Keywhenever the prompt, reference image, or parameters change for a new asset.
Contracts and related questions
Before sending a paid request, inspect the model contracts for free: model:gpt-6-luna, model:gpt-image-2.5-flare-std.

