Set up preventive maintenance

Goal: machines get serviced before they fail, on time, on usage, or on condition, with the work recorded so it can be measured.

Start with config_export. You'll change taskTypes and machineData for the machines in scope. Read guide_get topic maintenance.

1. The work itself

A maintenance task type per kind of service, for example "Laser monthly service" or "Press 1000-stroke service". For each:

a form with what the technician checks and records: checkboxes for each step, numbers for readings, a photo where it helps;
requiresFailureCoding: true on corrective types (repairs), so every repair records a problem, a cause and a remedy. Leave it off for routine PM;
the team or required skill, and SLAs per priority;
ssuStatusOnOpen / ssuStatusOnComplete if the machine's status should show it's under maintenance.

Trigger	Where in the spec	Example
Time	`recurrenceHours` on the task type: completing one schedules the next	Every 720 h (30 days)
Usage	`machineData[].meters` (manual) or a counter signal with `isMeter: true`, plus a `conditionRules` entry with `mode: "meter_delta"`	Every 1000 strokes, every 250 run hours
Condition	`machineData[].signals` plus a `conditionRules` entry with `operator` `gt` or `lt` and a `threshold`	Spindle temperature above 80 °C

Rules name a task type and a cooldown.
A rule never stacks tasks: while its task is open, it doesn't open another.
Ask the user for the intervals and limits. Don't invent them: they come from the machine's manual or the plant's practice.

machineData[].spareParts lists the parts each machine uses (SKU and quantity), so the parts for a service can be checked before it's due.

config_validate, then config_plan.
Show the user, per machine: the trigger, the interval or limit, and the task type it opens.
After their go-ahead, config_apply, then re-plan to confirm.

Technicians enter manual meter readings in the app, or you do with meters_record_reading from what they tell you. Its preview says whether it opens a task.
analytics_reliability (MTBF and MTTR) and analytics_failure_pareto show whether it's working.