[Bug]: Unable to pass image in request to Mistral models via Async Client

[ParthivNaresh]

What happened?

Starting with 1.69.1, I'm unable to include an image input into a request for Mistral models. This works on previous versions without any issues and I've confirmed the issue exists in 1.72.2. It also doesn't seem to extend to other model families. Would appreciate any help on this, thanks!

req = {
        'messages': 
        [
            {
                'content': 
                [
                      {
                          'text': 'Was an image passed in, yes or no?',
                          'type': 'text'
                      },
                      {
                          'image_url': {
                              'url': 'data:image/png;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7',
                              'detail': 'auto'
                          },
                          'type': 'image_url'
                      }
                    ],
                'role': 'user'
            }
        ],
        'model': 'mistral-medium-2505',
        'n': 1,
        'custom_llm_provider': 'mistral',
        'max_completion_tokens': 256,
        'api_key': 'XXXXXX'
    }

await litellm.acompletion(**req)

Relevant log output

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
File .../litellm/main.py:508, in acompletion(model, messages, functions, function_call, timeout, temperature, top_p, n, stream, stream_options, stop, max_tokens, max_completion_tokens, modalities, prediction, audio, presence_penalty, frequency_penalty, logit_bias, user, response_format, seed, tools, tool_choice, parallel_tool_calls, logprobs, top_logprobs, deployment_id, reasoning_effort, base_url, api_version, api_key, model_list, extra_headers, thinking, **kwargs)
    507 elif asyncio.iscoroutine(init_response):
--> 508     response = await init_response
    509 else:

File .../litellm/llms/openai/openai.py:765, in OpenAIChatCompletion.acompletion(self, messages, optional_params, litellm_params, provider_config, model, model_response, logging_obj, timeout, api_key, api_base, api_version, organization, client, max_retries, headers, drop_params, stream_options, fake_stream)
    764 response = None
--> 765 data = await provider_config.async_transform_request(
    766     model=model,
    767     messages=messages,
    768     optional_params=optional_params,
    769     litellm_params=litellm_params,
    770     headers=headers or {},
    771 )
    772 for _ in range(
    773     2
    774 ):  # if call fails due to alternating messages, retry with reformatted message

.../litellm/llms/openai/chat/gpt_transformation.py:400, in OpenAIGPTConfig.async_transform_request(self, model, messages, optional_params, litellm_params, headers)
    392 async def async_transform_request(
    393     self,
    394     model: str,
   (...)    398     headers: dict,
    399 ) -> dict:
--> 400     transformed_messages = await self._transform_messages(
    401         messages=messages, model=model, is_async=True
    402     )
    404     return {
    405         "model": model,
    406         "messages": transformed_messages,
    407         **optional_params,
    408     }

TypeError: object list can't be used in 'await' expression

Are you a ML Ops Team?

Yes

What LiteLLM version are you on ?

v1.72.2

Twitter / LinkedIn details

No response

[AmineDjeghri]

I am facing a similar issue with async mistral with text? Does it work for you using sync client instead of async ?

#11591

[ParthivNaresh]

I'm not seeing the error with async mistral with text, but it works with the sync client (text and image)

[AmineDjeghri]

okey, so i have a similar issue i think, i am using azure ai models (mistral, deepseek) and async calls don't work while sync calls do work