Physically buzz the user's phone with an urgent push notification.

The user is NOT watching the screen. Calling this interrupts them in the real world, so use it only when that interruption is justified.

USE FOR:

• a production alert / pager condition fired

• a long-running script, deploy, or migration FAILED or crashed

• a watched value crossed a critical threshold

• a task the user explicitly asked to be alerted about finished or blocked

• anything time-sensitive the user would want to know NOW, not later

DO NOT USE FOR:

• routine progress updates or status the user can read whenever

• ordinary task completion that is not time-sensitive

• anything the user did not ask to be interrupted about

• chatty / informational messages — those belong in the normal reply

Args: message: One-sentence body, shown on the lock screen. Be specific and actionable, e.g. "QA deploy failed: migration 0042 errored on playground_management" — not "something went wrong". severity: "info" | "warn" | "critical". - "critical" = loudest + most intrusive. May place a ringing phone call (SIP / Linphone — full-screen, pierces silent) and/or a Pushover emergency alert that repeats until acknowledged. Reserve for things that genuinely must interrupt the user right now. - "warn" (default) = high-priority push, non-repeating, no call. - "info" = normal-priority push, no call. title: Short notification title (a few words).

Returns a status string describing what each backend did.

Send a low-priority TEST notification to verify connectivity.

Safe to call during setup: it always uses 'info' severity so it will not fire a loud or repeating critical alert. Use this to confirm the phone receives pushes before relying on alert_me for real conditions.

Report which alert channels are configured (no secrets revealed).

Useful for debugging setup: shows the active channel, whether each backend has its credentials, and the current rate-limit settings.