← Back to docs

OpenAI image archive

OpenAI image archive

Tools stores generated images in a private archive when database image storage is enabled. Ordinary generated images, their prompts, and their reference/source material are private by default.

Public archive

The public archive is available at:

  • /openai/images/archive

Only images explicitly marked public are listed there. The archive can be filtered by user ID, category, and search text. Published images are grouped by user ID and, when one is set, the user's own category.

If an image has no category, it remains grouped under its user ID without requiring a category name.

Continue editing an archived image

Each image in the private archive has a Continue editing action. It opens a follow-up view where the archived image is attached as reference {1} for the next image generation.

This lets a nearly finished image be corrected without downloading and uploading it again. For example, the instruction can ask to correct a name or another small detail in {1} while preserving the rest of the image.

The new result is stored in the archive as usual. The result view includes Continue editing this result, so the same image chain can be continued through multiple revisions.

Signed-in users can also choose Use as reference on images in the public archive. A public image can be reused as an image reference by another user, but private prompts and private source material are not copied into the follow-up. Only the published image itself is reused.

Private archive images still follow ownership rules: they can be reused by the owner or an administrator.

Follow-up provenance records the previous generated image UUID as its origin, so archive-derived references are not recorded as ordinary file uploads.

Publishing your images

Signed-in users can open the private image archive at:

  • /admin/openai/images/archive

From there, Manage publishing opens the publication manager:

  • /admin/openai/images/archive/publishing

The publication manager supports selecting one image or many images on the current page and applying publication settings in bulk.

For selected images you can independently choose whether to:

  • publish or unpublish the generated image
  • publish or hide the stored prompt
  • publish or hide reference/source material
  • set, clear, or keep a category

Leaving a field as Leave unchanged preserves its current value.

X-bot images

Images created by the X-bot use special publication defaults:

  • generated image: public
  • prompt: public
  • references/sources: public
  • category: x-bot

These defaults also apply to previously archived X-bot images when the publication migration is installed.

Normal users cannot remove the X-bot publication defaults through the bulk manager. Administrators can override any publication field, including X-bot images.

Prompt and source privacy

Prompt and reference publication are separate from image publication.

A public image does not automatically expose its prompt or source material unless those visibility settings are also enabled. The public archive only shows user-facing prompt/source information. Internal system context, internal provenance metadata, hashes, and other operator-only details remain outside the public archive.

Reference image bytes are publicly readable only when both the generated image and its reference/source visibility are public.

Direct image URLs

Stored image URLs under /media/generated-images/{uuid} now follow the individual image's visibility state.

  • Public images can be read without signing in.
  • Private images require the owning signed-in account, an authenticated token belonging to the owner, or an administrator.
  • Stored reference images follow the same ownership rule unless the parent image and its reference/source visibility are public.

This means an unlisted UUID is not treated as a public link by itself.