LiteLLM ignores LLMRequest.model in generate_content_async

[markwaddle]

Describe the bug
google.adk.models.LiteLLM.generate_content_async ignores the model specified by the llm_request: LLMRequest, always using self.model. this prevents scenarios where an agent callback decides to use a specific model, and updates the llm_request.model, for example.

To Reproduce
Steps to reproduce the behavior:

1. declare an LLMAgent, specifying a LiteLLM as the model, passing in a specific model name. ex: LLMAgent(..., model=LiteLLM(model="openai/gpt5"), ...)
2. in an agent callback, for example before_model_callback, update the llm_request.model to a different value. ex: llm_request.model = "openai/gpt-nano"
3. invoke the agent

Expected behavior
LiteLLM should use the model provided by the llm_request. instead, it is fixed to the model it was initialized with.

Model Information:

• Are you using LiteLLM: Yes

Related line of code:

adk-python/src/google/adk/models/lite_llm.py

Line 821 in 29968d4

"model": self.model,

[ramanmangla]

Looking into this issue @GWeale