How does conductor keep the outputs of task, especially when the task in a workflow is completed?

Follow-up: Does it keep the whole output for that task, or only keep the one that is used on subsequence task or final output?

Conductor will store the input and output payloads for the tasks in the execution data store (redis, postgres, mysql…) this data is available to subsequent tasks in the workflow or after execution is completed for traceability. External payload storage is mainly to reduce pressure on the data store. So it’s worth looking at what you’re expecting in terms of payload size.