Image generation call replay incorrectly requires result and status

Source reports

• https://github.com/openai/openai-python/issues/2648

Current specification gap

Affected surface: POST /responses input; ImageGenToolCall via Item.

The source reports that the documented image-ID multi-turn example works while request typing requires result and status. The current public ImageGenToolCall requires both fields and reuses that component in input. This is distinct from assistant-message content issues #475 and #565.

Specification reviewed at b61ced9.

Minimal shape

{
  "type": "image_generation_call",
  "id": "ig_example"
}

This is a sanitized illustration of the reported shape, not a new live API capture.

Expected reconciliation

Model input replay separately from generated output as needed so an image ID can be reused without invented result/status values.

Verification scope

Reviewed the source reports, relevant comments, and current published schema. No new live API request was made. This mirrors the specification gap and preserves the original reproduction; it does not claim that a fix has shipped.

Start in openapi.yaml around line 42240 and compare the ImageGenToolCall component with its use for POST /responses input. Review the linked source report and comments, then separate input replay from generated output so the minimal type-and-id shape is accepted without result or status; preserve those fields where output requires them.