**`image.edit` fails with provider errors: `ReadError` and fal.ai `/image-to-image` 404**

Developers
1

Hi Anna team,

I’m testing a future cosmetic customization feature for Notebuddy: adding accessories to an existing character image while preserving its appearance.

The tests use a separate, installed private app:

  • App: notebuddy-image-edit-check
  • App ID: 519
  • Environment: Hosted Anna App runtime, using the App SDK
  • Time: October 12, 2026, approximately 04:58–05:01 UTC
  • The manifest permits image.edit, and the app’s image grant has enabled: true and allowEdit: true.

I made three requests, each requesting one image:

Test Result
Public GitHub reference image, with a GPT Image 2 model hint APP_PROVIDER_ERROR: image edit failed: ReadError:
The same image hosted in the Anna app bundle, with the same hint Same ReadError
Anna-hosted image, no model hint or optional rendering settings Error below

The third request returned:

APP_PROVIDER_ERROR
image edit failed: Exception: fal.ai image API call failed (HTTP 404):
{"detail":"Path /image-to-image not found"}

Minimal reproduction for the third request:

await anna.image.edit(
  {
    image_url:
      "https://anna.partners/anna-apps/kdkrkwhr/notebuddy/0.3.4/assets/starters-v1/ghost_water_stage1.png",
    prompt:
      "Add only one small purple wizard hat with a gold star to this blue water ghost. Preserve its face, eyes, smile, water markings, colors, pose, two wisp arms and curled vapor tail. It has no legs or feet: do not add any. Preserve the original painted illustration style and white background. Do not change anything else.",
    n: 1
  },
  { timeoutMs: 240000 }
);

Both source URLs returned HTTP 200 and identical image bytes when checked from our test environment. All three edit requests failed within approximately 1.9–2.5 seconds, without returning an image. The error responses did not identify the resolved model, so I cannot confirm which model handled the hinted requests.

Could you help clarify:

  1. Which model/provider route currently supports image.edit, and how should we select it?
  2. Does the /image-to-image 404 indicate a provider routing/configuration issue, or does our request need to change?
  3. Can these failed requests consume Anna allowance?

These were isolated image-edit tests; the production app and companion data were not modified. We have stopped further requests pending clarification.

Thank you!