Alerts

SQL-native alert rules that evaluate queries and fire when conditions are met. The functions return rows; wrap them in your scheduler of choice (pg_cron, Databricks Jobs, etc.) to drive notifications.

alert_check(p_name, p_query, p_severity)

Runs a query and returns any "firing" rows as alert events.

Parameter	Type	Default	Description
p_name	TEXT	—	Alert rule name
p_query	TEXT	—	Query that returns rows when the alert should fire
p_severity	TEXT	'warning'	Severity: 'info', 'warning', 'critical'

Returns: TABLE — alert_name, severity, fired_at, alert_data (JSONB)

-- Fire when any device has avg temperature > 80°C in the last hour
SELECT * FROM lakets.alert_check(
    'high_temperature',
    $$SELECT device_id, avg(temperature) AS avg_temp
       FROM sensor_data
       WHERE time >= now() - interval '1 hour'
       GROUP BY device_id
       HAVING avg(temperature) > 80$$,
    'critical'
);

alert_deadman(p_name, p_table_name, p_group_by, p_timeout, p_schema_name)

Detects silent series — groups that haven't reported data within the timeout. Essential for monitoring data-pipeline health.

Parameter	Type	Default	Description
p_name	TEXT	—	Alert name
p_table_name	TEXT	—	ChronoTable to check
p_group_by	TEXT	—	Column to group by (e.g. 'device_id')
p_timeout	INTERVAL	—	Maximum silence before alerting
p_schema_name	TEXT	'public'	Schema

Returns: TABLE — alert_name, severity (always 'critical' for deadman alerts), fired_at, dead_key, last_seen, silent_for

-- Alert if any device hasn't reported in 15 minutes
SELECT * FROM lakets.alert_deadman(
    'device_heartbeat', 'sensor_data', 'device_id', '15 minutes'
);
-- device_heartbeat | critical | 2026-04-09 10:30 | device-17 | 2026-04-09 10:12 | 00:18:00