Advanced

These settings have sensible defaults and most houses never need to touch them. Two of them protect your server's storage, one decides how soon a device is named a repeat offender, and one sets how long your maintenance window stays open.

Two settings on this page, the storage write interval and history retention, control your server's storage footprint. If you run Home Assistant from an SD card, turning down the write frequency and keeping less history will protect your hardware from burnout. If you run on a fast mini-PC with an SSD, you can keep the defaults or tune them for maximum detail. Built to Be Efficient covers the economics of these choices in detail.

Repeat Offender Threshold

Your Daily Brief carries a table called Repeat Offenders for devices whose interruptions keep having no clear explanation. Because the system can't guess why a TV goes unavailable when turned off or why a dying coin cell triggers repeatedly, it names the pattern and lets you decide.

An interruption counts toward this table only when nothing explains it: no restart, bridge or broker outage, integration reload, or maintenance window.

This slider controls how many unexplained interruptions inside a week put a device in the table (default is 2).

• Lower it: Devices are named sooner, on weaker patterns.
• Raise it: The table waits for a stronger, more persistent pattern before naming anything.

How Much History to Keep

Device Sentinel keeps a concise daily record per device (longest gap, signal readings, battery level) to help you spot slow-moving trends over time, like battery decay or seasonal radio degradation.

The setting ranges from 30 to 360 days (default is 90 days).

• On an SD Card: Drop this to 30 days to keep your storage footprint microscopic.
• On an SSD: Leave it at 90 or scale it up if you love long-term charts.

Lowering this setting takes effect at the next midnight, not the moment you move the slider, so a mistake can be undone before any history is lost.

A device that stays silent for longer than the history you keep is recorded as silent for that long, and shown as "more than 90 days" (or whatever you keep). Device Sentinel cannot measure a silence longer than its own memory, so it says so rather than guessing.

Storage Write Interval

Routine activity clocks are held in memory and flushed to disk on a schedule rather than every time a sensor blinks (default is 15 minutes).

Important events, like a new problem, a recovery, or an acknowledgment, are always written immediately.

• On an SD Card: Crank this up to 30 or 60 minutes to maximize card longevity.
• On an SSD: Leave it at 15 (or drop to 5) to narrow the tiny window of routine clock progress exposed to a sudden power cut.

Maintenance Window in Minutes

When you manually service hardware, you press the Maintenance Mode button so your repair isn't mistakenly learned as normal downtime. This slider sets how long that maintenance window stays open once pressed (default is 10 minutes).

Data Trim

At the foot of the screen sit the data trim pickers. Use these only when support asks you to, or when you know a device's record has gone wrong and you want it to start over from scratch.

What it erases: A device's statistics, silence episodes, incidents, and Problem List entries.

What survives: The device itself (which is rediscovered instantly and learned again from the start: about a week before it can be called frozen, and two weeks before its battery can be judged falling), all mutes, all exclusions, the battery, power and type answers you set with the pencils, and the house-wide record of restarts and outages.

The safeguards: Before anything touches deletion, an automatic backup of your storage files is safely stashed away in config/device_sentinel/trim_backups. If that backup cannot be written, nothing is deleted. The pickers also empty themselves after the save, so saving another setting later cannot repeat the erasure.