LLM output not logged to Langfuse tracer when using AsyncPipeline

[henryclw]

Describe the bug
When I use AsyncPipeline with Langfuse tracer, the llm output (i.e. the replies) is not logged to Langfuse

This would work, both input and output of the llm are logged:

    tradition_pipeline = Pipeline()
    tradition_pipeline.add_component("tracer", LangfuseConnector("haystack-pipeline-testing"))
    tradition_pipeline.add_component("llm", OpenAIChatGenerator())

This would won't work, only the input of llm is logged:

    async_pipeline = AsyncPipeline()
    async_pipeline .add_component("tracer", LangfuseConnector("haystack-pipeline-testing"))
    async_pipeline .add_component("llm", OpenAIChatGenerator())

FAQ Check

• Have you had a look at our new FAQ page?

[henryclw]

@sjrl Thank you for the fix. That's awesome.