When using this API many times in a row to try and populate real policy
descriptions/resolutions, I discovered a problem: the inferred
resolution (“What we’ll do”) almost never used decisive language, so
you'd get a lot of “probably”, etc., when we really want it to sound
more confident about what will happen during the maintenance window.
(The calendar events say what WILL happen to your device, not what
should probably happen.)
After several attempts to tweak the prompt, I ended up getting the best
results by just removing "probably" from
`whatWillProbablyHappenDuringMaintenance`. I made this change to the
prompt without changing the data that this API returns.
---------
Co-authored-by: Eric <[email protected]>
+ Use the same "Failed to generate human interpretation using generative
AI." message if you send it something weird like "asdf" (so you don't
get a vague 500 error).
+ Simplify the code a bit
---------
Co-authored-by: Eric <[email protected]>
