Synthesized alerts — `alerts`

Evaluates a PromQL ratio each poll and emits an alert event when it exceeds a threshold. Use it to put a “this is bad” marker in the timeline without running Alertmanager.

Prerequisites: a reachable Prometheus server.

Set up

Agent

Ask a connected agent:

Use navflow to alert when the 5xx percentage on localhost:9090 exceeds 5.

The agent composes the ratio expression, test_source dry-runs it, then create_source.

Console

Sources → Add source → alerts. Set:

url: the Prometheus base URL
ratio_promql: the expression returning the ratio to watch
threshold: fire above this value

Add a const label naming the service (mark it primary). Save.

YAML / API


- name: alerts
  connector: alerts
  poll: 5s
  config:
    url: http://localhost:9090
    threshold: 5
    ratio_promql: '100*sum(rate(http_requests_total{status=~"5.."}[1m]))/sum(rate(http_requests_total[1m]))'
    labels:
      - { name: service, const: api-server, primary: true }

Or POST /api/sources with the same object.

Configuration

field	type	description
`url`*	string	Prometheus base URL
`ratio_promql`*	string	PromQL expression returning the ratio to watch
`threshold`	number	fire when the ratio exceeds this value (default 5)
`key`	string	legacy fixed key — prefer a primary `const` label

An alerts source marks the timeline; a trigger pushes. To wake an agent when a condition holds, attach a trigger to a view instead — see Watch with triggers.

Verify

Push the ratio over the threshold (or lower threshold temporarily) and watch the alert event land on the next poll.

Synthesized alerts — alerts