FLUX.1-Kontext-pro model and Azure AI Foundry.

I deployed the FLUX.1-Kontext-pro model to Azure AI Foundry.

While the image generation works great, the editing endpoint does not seem to work.

Here is the error that is returned:

Failed to generate image: No model specified in request. Please provide a model name in the request body or as a x-ms-model-mesh-model-name header. Trace ID : 6199947b-82fa-4314-98db-03d314a3afc1

I used the same code provided in the playground. Trying to edit in the playground gives the same error message.

Does the edit endpoint work at all ?

# Initialize the DefaultAzureCredential to be used for Entra ID authentication.
# If you receive a `PermissionDenied` error, be sure that you run `az login` in your terminal
# and that you have the correct permissions to access the resource.
# Learn more about necessary permissions:  https://aka.ms/azure-openai-roles
credential = DefaultAzureCredential()
token_response = credential.get_token("https://cognitiveservices.azure.com/.default")

base_path = f'openai/deployments/{deployment}/images'
params = f'?api-version={api_version}'

generation_url = f"https://xxxxxx.cognitiveservices.azure.com/{base_path}/generations{params}"
generation_body = {
  "prompt": "create a image of cats on a playground",
  "n": 1,
  "size": "1024x1024",
  "output_format": "png"
}
generation_response = requests.post(
  generation_url,
  headers={
    'Authorization': 'Bearer ' + token_response.token,
    'Content-Type': 'application/json',
  },
  json=generation_body
).json()
save_response(generation_response, "generated_image")

# In addition to generating images, you can edit them.
edit_url = f"{endpoint}{base_path}/edits{params}"
edit_body = {
  "prompt": "add dogs to the playground",
  "n": 1,
  "size": "1024x1024"
}
files = {
  "image": ("generated_image.png", open("generated_image.png", "rb"), "image/png"),
}
edit_response = requests.post(
  edit_url,
  headers={'Authorization': 'Bearer ' + token_response.token},
  data=edit_body,
  files=files
).json()
save_response(edit_response, "edited_image")