Duplicate langfuse trace if use litellm.acompletion in CustomLLM.acompletion

[Yuhui0620]

I try to use litellm.acompletion sdk to request llm in the overrided method CustomLLM.acompletion like that:

class MyCustomLLM(CustomLLM):
    async def acompletion(
        self,
        model: str,
        messages: list,
        api_base: str,
        api_key,
        optional_params: dict,
        litellm_params=None,
        **kwargs
    ) -> ModelResponse:
        ...
        return await litellm.acompletion(
            model=f"openai/{model}",
            messages=messages,
            base_url=api_base,
            api_key=api_key,
            metadata=metadata,
            **optional_params
       )

But a duplicate trace will be generated.

Is there any way to skip the second langfuse callback?