peal.preflight.check_config¶
- peal.preflight.check_config(path, check_existence=True, will_exist=None)[source]¶
Return (problems, notes) for one config. Problems mean it cannot run.
With
check_existence=Falseonly the machine-independent checks run (parses, no conflict markers, no foreign paths), which is what a commit hook on a laptop without$PEAL_RUNScan honestly assert.will_existholds artefacts that an earlier line of the same reproduction script creates. The scripts are ordered: they train a generator and then run the explainers that consume it, so an input that is absent today but produced upstream is not a problem, it is simply a step that has not been run yet. Without this the checker reports a script as broken when it is merely unrun.