re: “how can I specify the parent_run_id that should be used for these upstream dependencies? Is the default behavior to use the previous successful run’s outputs?”

execute_pipeline()

doesn’t take parent_run_id bc the run created by this api will be treated as a root run, i.e. a brand new run rather than a child run of any previous run. there’s two ways to “re-execute” a pipeline when you want o skip some upstream solids and use the previous intermediates) 1. use the

reexecute_pipeline

but for now you need to list all the steps that you wanna run bc it currently won’t resolve the query language (we are working on it) example: https://docs.dagster.io/tutorial/advanced_intermediates#reexecute-a-pipeline-through-python-api 2. or you can

execute_pipeline

and specify the location of the previously stored intermediates in

run_config

the run_config would be something like https://github.com/dagster-io/dagster/blob/master/examples/docs_snippets/docs_snippets/intro_tutorial/advanced/intermediates/reexecution_env.yaml but in this case, the system will treat the run as a brand new root run

you will lose the parent-child run lineage