mime_types trucates "image/", passing wrong mime types such as "jpeg"/"png"

[SaarYogev]

** Please make sure you read the contribution guide and file the issues in the right place. **
Contribution guide. ✅

Describe the bug
When sending parts with a blob containing an image using mime_type and data, the mime_type gets transformed from a type image/jpeg to jpeg.
This then causes the LLM call to fail, as the mime_type is not on the acceptables mime_types list, as this list linked in the error shows:
https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/inference

To Reproduce
Steps to reproduce the behavior:

1. Install 'google-adk==1.7.0'
2. Run parts.append(types.Part(inline_data=types.Blob(mime_type="image/jpeg", data=image_data_base64)))
3. Send the request to Gemini 2.5 Pro via Vertex AI
4. See error:

litellm.exceptions.BadRequestError: litellm.BadRequestError: VertexAIException BadRequestError - {
  "error": {
    "code": 400,
    "message": "Unable to submit request because it has a mimeType parameter with value jpeg, which is not supported. Update the mimeType and try again. Learn more: https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/gemini",
    "status": "INVALID_ARGUMENT"
  }
}

Expected behavior
A 200-code response from the LLM, reading the image data correctly, just like it did with google-adk 1.6.1.

Desktop (please complete the following information):

• OS: [e.g. iOS] Ubuntu 22.04
• Python version(python -V): Python 3.12.3
• ADK version(pip show google-adk): 1.7.0

Model Information:
Gemini 2.5 Pro

Additional context
I've located the exact PR and line which cause the issue:
https://github.com/google/adk-python/pull/1917/files
lite_llm.py:256
Why was it added I do not know, but I hope an existing contributor or the PR author himself would.

[Shishir420-GIT]

'Invalid argument' is probably, I was just checking some docs, don't really see the need to the additional formatting,

The mime type is being truncated so instead of 'image/jpeg' which is correct I believe, 'jpeg' is getting sent, the full MIME type should already be preserved in the data_uri as 'data:image/jpeg;base64'

Simply removing the separate format field eliminates the truncated MIME type in both video and image, (from Line 254 in lite_llm.py) so it should resolve the issue.

Like this:

  # Before:
  if part.inline_data.mime_type.startswith("image"):
      # Extract format from mime type (e.g., "image/png" -> "png")
      format_type = part.inline_data.mime_type.split("/")[-1]
      content_objects.append(
          ChatCompletionImageObject(
              type="image_url",
              image_url=ChatCompletionImageUrlObject(
                  url=data_uri, format=format_type
              ),
          )
      )

  # After:
  if part.inline_data.mime_type.startswith("image"):
      content_objects.append(
          ChatCompletionImageObject(
              type="image_url",
              image_url=ChatCompletionImageUrlObject(url=data_uri),
          )
      )

and same for video as well.

[ajai-abilytics]

+1

[polong-lin]

@Jacksunwei Could you please take a look at the proposed fix in #2070 (comment)?

[Shishir420-GIT]

@polong-lin let me know, if you'd like any additional support on this.

[hborras]

Same issues here.

[srochiramani]

or help approve this -> #2612

[hborras]

or help approve this -> #2612

I don't think that would be the solution, because with openAI models it is working. You can't just delete the code

[hborras]

Confirming it doesn't work on Gemini or Claude models loaded throught litellm, but is working with OpenAI models

[SaarYogev]

What's the status of this issue @Jacksunwei ? I see it got stuck in "FixIt Week" then didn't move.
An update on the priority of this issue and consequently when should a fix be expected would be very appreciated.

[Shishir420-GIT]

What's the status of this issue @Jacksunwei ? I see it got stuck in "FixIt Week" then didn't move. An update on the priority of this issue and consequently when should a fix be expected would be very appreciated.

It seems to have been merged under PR: #2612