feat(pipelines): support multiple streaming-capable components

[MrCatters]

Previously the pipeline utilities expected exactly one streaming component. The new implementation allows any number of components to define a streaming_callback attribute:

• find_streaming_components() returns a list of tuples containing streaming_component, streaming_component_name
• _setup_streaming_callback_for_pipeline() iterates over all streaming components and attaches the provided callback
• _validate_async_streaming_support() checks every streaming component for async capabilities and reports all failures in a single error message

[MrCatters]

tests/test_it_pipeline_utils.py will need to be updated as well. I try not to mess with testcases on projects I am not familiar with and typically leave it up to the maintainers 😀.

[hoschmieder]

Hi, MrCatters,
do you address the same issue as in my request?
#133

[mpangrazzi]

@MrCatters @hoschmieder Can I ask you about the use case of this PR?

[hoschmieder]

Hi @mpangrazzi,
Use case of this (PR: 133):
I have multiple compontens within a pipeline, e.g. two llm's, vector store ...
all of the providing streaming output to inform the user about the process.
with the original code i just can stream out one component.
with my patch each component can stream in realtime.
example within a own component:

        if streaming_callback:
            streaming_callback(StreamingChunk(content=f"⏳ Der Ranker is starting ... it takes a while">

Maybe there are better way's to solve this. But that works for my demands today.