What happened?

Trying to use partial image loading with image generations like:

  stream = client.images.generate(
      model="gpt-image-1",
      prompt="A cute sea otter",
      n=1,
      size="1024x1024",
      stream=True
      partial_images=2
  )

  for event in stream:
      if event.type == "image_generation.partial_image":
          idx = event.partial_image_index
          image_base64 = event.b64_json
          image_bytes = base64.b64decode(image_base64)
          with open(f"river{idx}.png", "wb") as f:
              f.write(image_bytes)

Yields this:

openai.BadRequestError: Error code: 400 - {'error': {'message': 'litellm.BadRequestError: AzureException BadRequestError - {\n  "error": {\n    "message": "Unknown parameter: \'extra_body\'.",\n    "type": "invalid_request_error",\n    "param": "extra_body",\n    "code": "unknown_parameter"\n  }\n}. Received Model Group=openai_gpt_image_1\nAvailable Model Group Fallbacks=None', 'type': None, 'param': None, 'code': '400'}}

This is because the request body sent to Azure has the field extra_body, like:

{'model': 'gpt-image-1', 'prompt': 'A cute sea otter', 'n': 1, 'size': '1024x1024', 'extra_body': {'partial_images': 2, 'stream': True}}

(look at the variable data at litellm/llms/azure/azure.py:1130)

but Azure doesn't know what extra_body is and expects:

{'model': 'gpt-image-1', 'prompt': 'A cute sea otter', 'n': 1, 'size': '1024x1024', 'partial_images': 2, 'stream': True}

This request body is originally created in get_optional_params_image_gen() at litellm/utils.py:2535.

So the data should probably be formatted and flattened correctly in get_optional_params_image_gen() (or at a later point). But I unfortunately don't have the time at the moment to actually understand what get_optional_params_image_gen() does well enough to suggest an actual change, will leave that up to you!

What LiteLLM version are you on ?

v1.77.5

Twitter / LinkedIn details

No response