Chris works from a list of the services I expect to be running. It checks them regularly and notices when a result changes. It can also flag an unfamiliar application running on the machine for me to review.

The list gives the tool a specific job, and it gives me a baseline when a machine starts behaving differently from what I expect.

One failed check doesn't always call for a restart. It may be a brief interruption, so Chris checks again before acting and then uses only the repairs I've allowed for that particular service. For some applications, that means a restart followed by another check.

Storage, network access and important database services stay outside the automatic repair routine; those need my attention. If a problem persists, the process stops and reports. Repeating the same repair forever would create more work than the original fault.

The AI component writes the explanation: which service failed or recovered and what the watchdog tried. The checks and repair rules do the operational work. The model describes the results; it doesn't choose or execute repairs.

Chris compares each result with the previous one, so repeated checks don't have to become repeated alerts. A daily briefing collects the previous day's application checks, repairs, machine activity and selected traffic information, distinguishing a service that recovered from one that still needs attention.