Skip to main content

Reliability

The Reliability screen shows final client-visible service failures for the selected period. The overall rate is useful for orientation, but the actionable view is the breakdown enabled for the organisation.

Corporate reliability commitment​

OmevAI's standard corporate reliability commitment is no more than 1% final client-visible service failures for each agreed work category separately. Latency, RPM and TPM targets are agreed individually with the organisation.

A final client-visible service failure means that a request ultimately ended in a service error or returned no usable API response. An internal attempt recovered before the final response is not counted as a client-visible failure.

If an agreed output specification includes a machine-checkable format, failure of that formal check is included only when this has been expressly agreed. The business and factual quality of answer content is reviewed separately; an HTTP 200 does not by itself prove that the answer meets the intended use.

Reading the screen​

  • Failure share is final failed requests divided by all completed requests in the selected period.
  • Breakdown shows the same ratio for each visible category, so one problem is not hidden inside a healthy average.
  • Request export provides status and latency for detailed reconciliation.

Client input rejected before processing is not automatically a service failure. Check the HTTP status and the organisation's measurement rules.

When numbers disagree​

First compare the same completed period and timezone, then compare requests with requests. Recent minutes may still be arriving. If the difference remains, send support@omev.be the period, affected category, request identifiers, key alias or mask, and business impact.