You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
modular flows: flows should be structured such that only blocking tasks are in a single flow (dynamic or otherwise)
e.g.: we should decouple returning solutions from the grading flow, because a failure to autograde shouldn't prevent solution return
make-like tasks: if the datestamp of a precursor is newer than the result, re-process
this will allow things like after-the-fact extensions, regrading, changing manual grading, etc
implementation agnostic: rudaux should only know about high-level abstractions (LMS, student storage, etc), not implementations (Canvas, ZFS over SSH, etc)
sysadmin-minimal: the teaching team should have to use the linux command line for as little as possible (except starting rudaux and specifying basic course configuration)
e.g. the instructor should be able to specify graders, assignments to grade, etc via the LMS
We can fill in more here as we think of them in the initial design phase of the new version of Rudaux
ability to re-assign graders after the fact?
pydantic for data validation (not traitlets)
The text was updated successfully, but these errors were encountered:
trevorcampbell
changed the title
Design principles for upcoming Prefect Orion version
Design for upcoming Prefect Orion version
Jun 3, 2022
We can fill in more here as we think of them in the initial design phase of the new version of Rudaux
The text was updated successfully, but these errors were encountered: