Vaktmester monitors the automations running on my Mac. It checks recent runs, failure signals and the outputs each job is expected to produce.
Checking the result
The success condition depends on the job. For a publication it can be a dated delivery record; for a sync it can be a completed collection. A recent log or a successful process exit is not sufficient on its own.
The monitor accounts for retry windows and temporary network failures. It groups new problems and suppresses repeated alerts, so an existing error does not generate the same message on every check.
State is associated with stable job names. This prevents a moved log file from being mistaken for a fresh collection of old errors.
The wider tooling records execution time and model usage for investigation. It is built in Python and uses local state files and macOS scheduling.
Related work: Internship Radar and study sync.