Measured uptime

Fallbacks

• Two application servers and three media servers share the calls. One media server is enough for calls to continue.
• We update one server at a time: it stops taking new calls and waits for the calls in progress to finish.
• If the language model provider does not respond, the call continues on a backup model. If speech synthesis fails, the agent continues with a backup voice provider.
• If a transfer cannot connect, the caller stays with the agent, which tells them or takes a message.
• Your phone system, edge device and carrier lines reconnect on their own as soon as the cause goes away. Also set a fallback route on your phone system (an extension, a ring group or voicemail), so callers still reach someone while the connection is down.

Alerts when your connection drops

• If your phone system or edge device has not connected for more than 10 minutes, we email you a "Phone System Offline" alert. It repeats every 4 hours while the connection stays down.
• The same applies to a line from your own carrier that Voice Logica connects. If the carrier refuses the login, we pause the retries so the account is not locked, and let you know.
• Emails go to your company’s alert addresses, or to the company email when none are set. No alert is sent outside your business hours. If many customers drop at the same moment, the cause is usually on our side: our own team is alerted.

Details: If a connection drops ↗

What we measure

Every 15 seconds we check that the platform API, the call gateway (SIP) and at least one media server respond. A check counts as passed only when all three respond.

When the monitoring system itself was not recording, that time counts neither as uptime nor as downtime. Percentages are rounded down.

Measurements are kept from 24 October 2026. Older measurements were not kept and are not shown.

What the checks cannot see

• Interruptions shorter than 15 seconds. Such a short gap falls between two checks and goes unrecorded. Before a restart, the media servers wait for the calls in progress to finish, and new calls keep being accepted.
• Incidents where the servers respond but calls do not go through properly. For example, on 3 October 2026, 08:41–09:22 UTC, agents went silent after the greeting while every check passed. We find such incidents through call analysis.
• Problems in the telephone carrier’s network or in the customer’s own phone system.