Input: live checks and logs
Analysis is fed from two sources: the results of live connectivity checks at the time of the incident, and syslog records from the same time window. Analysis is run by the operator, on a per-incident basis.
Output: summary, probable causes, steps
The output has three parts. The summary conveys verified findings. Probable causes are hypotheses that still need to be verified and are shown separately from findings. Troubleshooting steps are ordered suggestions, each noting the expected result.
This separation is not cosmetic. Presenting a hypothesis as a finding is the most common cause of a wrong intervention.
The decision boundary
The system never applies any step on its own. Because actions such as a restart can cause an outage, the decision belongs to the operator.
No definitive root cause is declared, and no outage prediction is promised. The confidence score shows how clear the signal behind an assessment is; it is not a guarantee of accuracy.
Feedback and memory
The operator's 'worked / did not work' mark is written into the analysis history. When a similar incident recurs, you can see what was tried before; institutional memory builds up this way.