Park every state

AnalyticsSDKHTTPDashboardconditionalsubtypeended_at

Each state an item passes through is one escalation row. The queue becomes a complete time series of the process with no extra instrumentation.

The situation

You want to know how machines spend their day: idle, running, waiting to be unloaded, in service. Adding timers and a metrics pipeline to every step is work that drifts from the process itself.

The recipe

const run = await conditional<RunResultV1>(`run-${serialNumber}-${cycle}`, {
  role: 'machine-fleet',
  type: 'fleet',
  subtype: 'running',
  metadata: { serialNumber, model: 'mx-200', coolant: false },
});

const unloaded = await conditional<{ ok: boolean }>(`unload-${serialNumber}-${cycle}`, {
  role: 'unload',
  metadata: { serialNumber, model: 'mx-200' },
});

role is the queue that attends the state. subtype names the state within a role that holds several. metadata carries the entity key as a string plus any categorical facts worth slicing by.

The guarantee

Every row is one interval [created_at, ended_at). ended_at is resolved_at for resolved rows and the guarded transition time for cancelled or expired rows, and stays null while the row is live. The workflow already writes everything the analytics read; there is nothing else to emit. See Analytics.

Analytics