I am using "@mastra/langfuse": "1.0.2" in combination with a Mastra agent using gpt-5.1 llm.
When I'm looking at traces in Langfuse, it does not show the input. I have to drill down to the llm trace to see the actual user input query. Also the output has the full object, not just the text.
Is there a way to configure the root observation to have the value Langfuse expects so it shows up correctly in the Langfuse dashboard?