Hi Team, how would you, or what is the ClearML recommended practice for organising sub-steps inside each pipeline step?

For example I have several sources of data to retrieve, each of which is potentially a sub step as I wanted to group them into one big pipeline step called "ingestion" ( I don't think make those sub step a pipeline step a good idea anyway).

What would be a good practice you recommend? Having class gathering them all will not record all the dependencies in each sub step correctly, right? (local imports are in each sub step's function code)

@PipelineDecorator.Component()
def ingestion_step():
    from somewhere.ingestion import Ingestion
    ingestion = Ingestion()
    ingestion.run()  # Runs all the sub step inside this class

class Ingestion:
  def __init__(self):
       self.state1 = 'Ah'
       self.a_data = None
       self.b_data = None
  
  def retrieve_a(self):
      # This import is not visible to ClearML?
      from sources.a import ASource
      self.a_data = ASource.retrieve()