[Bug]: Gemini model parameters not set by using GoogleAIStudioGeminiConfig

[mrm1001]

What happened?

In your docs, you show how to set the parameters of OpenAI in this way:

litellm.OpenAIConfig(max_tokens=10)

response_2 = litellm.completion(
            model="gpt-3.5-turbo",
            messages=[{ "content": "Hello, how are you?","role": "user"}],
        )

I'm trying to take this approach with Gemini:

import litellm

litellm.GoogleAIStudioGeminiConfig(temperature=0, top_k=3)

def my_custom_logging_fn(model_call_dict):
    print(f"model call details: {model_call_dict}\n\n")

response = litellm.completion(
    model="gemini/gemini-2.0-flash-exp", 
    messages=[{"role": "user", "content": "hi there"}],    
    logger_fn=my_custom_logging_fn
)

response

The logs that get printed do not have the parameters in them (abbreviated). The generationConfig dictionary is empty.

Final returned optional params: {}
[...]

model call details: {'litellm_trace_id': None, 'litellm_call_id': 'd74ca109-1042-480b-b62a-e0ea776c4a2d', 'input': [{'role': 'user', 'content': 'hi there'}], 'model': 'gemini-2.0-flash-exp', 'messages': [{'role': 'user', 'content': 'hi there'}], 'additional_args': {'complete_input_dict': {'contents': [{'role': 'user', 'parts': [{'text': 'hi there'}]}], 'generationConfig': {}}, 'api_base':

If I try passing the parameters directly, as such:

response = litellm.completion(
    model="gemini/gemini-2.0-flash-exp", 
    messages=[{"role": "user", "content": "hi there"}],    
    top_k=3,
    temperature=0,
    logger_fn=my_custom_logging_fn
)

Then the parameters do get passed to the endpoint (abbreviated logs):

Final returned optional params: {'temperature': 0, 'top_k': 3}
[...]

model call details: {'litellm_trace_id': None, 'litellm_call_id': 'd74ca109-1042-480b-b62a-e0ea776c4a2d', 'input': [{'role': 'user', 'content': 'hi there'}], 'model': 'gemini-2.0-flash-exp', 'messages': [{'role': 'user', 'content': 'hi there'}], 'additional_args': {'complete_input_dict': {'contents': [{'role': 'user', 'parts': [{'text': 'hi there'}]}], 'generationConfig': {'temperature': 0, 'top_k': 3}}, 'api_base':

Relevant log output

Are you a ML Ops Team?

No

What LiteLLM version are you on ?

v1.58.2

Twitter / LinkedIn details

No response