Self-monitoring

Monitor the health of Anaphora and the success of its jobs and deliveries.

Health monitoring

Access the monitoring settings at Settings > Application > Health Monitoring.

Set up health alerts

Get alerts when Anaphora detects changes in job success rates.

Under Operator notifications > Delivery, select a Delivery Interface. For an SMTP or Mailgun interface, also add the Recipients. A webhook with Define body in job instead is not available.
Select Job Health Alerts.
Set Health Check Frequency: how often to check job health.
Set Maximum Notification Frequency: minimum time between alerts (optional).
Optionally, click Test to send the current health status.
Click Save.

info

The same delivery settings also receive the token budget alerts of the AI providers.

Health API

To get the health status of Anaphora through the API, call this endpoint:

GET /guest/api/health

What the answer contains depends on the credentials that you send:

You send	You get
Nothing	The worst colour (`status`), the `counts` per colour, and one `healthStatus` per job and delivery interface. No names.
An observer key: `Authorization: Bearer <key>`	Every job and delivery interface by name, with its schedule, its recent runs and its delivery counts. No error texts.
The credentials of a system user: `Authorization: Basic …`	All of the above, and the delivery error texts of the last 24 hours.

The answer is always HTTP 200, except for an observer key that is revoked or mistyped: that gets HTTP 401, so your monitor raises an alert. Anaphora keys start with ana_. A bearer token that does not, for example the token of a proxy in front of Anaphora, gets the answer without names.

Observer keys

Use an observer key for a monitoring tool (Centreon, Zabbix, Nagios, Prometheus blackbox exporter). The key reads the health status only, and you can revoke it without changing a user's password.

Go to Settings > Application > API Keys.
Click New key. Give the key the name of the monitor that uses it, and click Create.
Copy the key. Anaphora shows it once and keeps only a hash of it.
To test the key, run the curl command that Anaphora shows under the key.
Configure the monitor to send the key in the Authorization header:

curl -fsS -H "Authorization: Bearer ana_obs_..." 'https://anaphora.example.com/guest/api/health'

With -f, curl fails with error 401 when the key is wrong, and does not print the answer without names.

To test a key later, use the command at the top of the API Keys page. It asks for the key, so the key does not go into the command or the shell history. Paste the key and press Enter. The key stays hidden:

read -rs ANAPHORA_API_KEY && curl -fsS -H "Authorization: Bearer $ANAPHORA_API_KEY" 'https://anaphora.example.com/guest/api/health'

The list shows when each key was last used. To revoke a key, click Revoke next to it, and confirm. From then on, a monitor that sends the key gets 401.

Send the key in the header only. Anaphora does not read a key from the query string, because query strings are written to proxy logs.

Response format

Without credentials:

{
  "status": "yellow",
  "counts": {
    "jobs": { "red": 0, "yellow": 1, "green": 1, "gray": 0 },
    "deliveryInterfaces": { "red": 0, "yellow": 0, "green": 1, "gray": 1 }
  },
  "jobs": [{ "healthStatus": "yellow" }, { "healthStatus": "green" }],
  "deliveryInterfaces": [{ "healthStatus": "green" }, { "healthStatus": "gray" }]
}

With an observer key, the answer also names every job and delivery interface. A system user also gets the summary24Hours.errors list of each delivery interface, which an observer key does not get:

{
  "status": "green",
  "counts": {
    "jobs": { "red": 0, "yellow": 0, "green": 1, "gray": 0 },
    "deliveryInterfaces": { "red": 0, "yellow": 0, "green": 1, "gray": 0 }
  },
  "jobs": [
    {
      "id": "79cf54b6-df32-4b09-84f4-708ecc72b7bc",
      "name": "Kibana Dashboard Snapshot",
      "description": "Takes a snapshot of a dashboard",
      "cron": "5 4 * * *",
      "healthStatus": "green",
      "recentRuns": [
        {
          "runAt": "2026-01-15T04:05:00.041Z",
          "state": "success"
        },
        {
          "runAt": "2026-01-14T04:05:00.051Z",
          "state": "success"
        },
        {
          "runAt": "2026-01-13T04:05:00.034Z",
          "state": "success"
        },
        {
          "runAt": "2026-01-12T04:05:00.040Z",
          "state": "success"
        },
        {
          "runAt": "2026-01-11T04:05:00.034Z",
          "state": "success"
        }
      ]
    }
  ],
  "deliveryInterfaces": [
    {
      "id": "4d5fba03-561e-4503-bf4d-c41817133aca",
      "name": "My Delivery Interface",
      "type": "webhook",
      "healthStatus": "green",
      "summary24Hours": {
        "totalCount": 11,
        "errorCount": 0,
        "errors": []
      }
    }
  ]
}

Health status values

Status	Meaning
green	All recent runs successful
yellow	Some recent failures
red	All recent runs failed
gray	No recent activity

Run states

Each entry of recentRuns has a state:

State	Meaning
`success`	The run succeeded and its report was delivered
`partial`	The report reached some delivery interfaces or recipients, but not all
`failed`	The capture failed, or the report reached nobody (for example a withheld report)

A job is green when its last five runs are all success, and red when they are all failed. A partial run makes the job yellow, not red: one address that refuses the mail keeps the job yellow.

AI budget alerts

The delivery interface of the health alerts also carries the AI token budget alerts. You can set it without switching health monitoring on.

Next steps

Backup: configure backups

Health monitoring​

Set up health alerts​

Health API​

Observer keys​

Response format​

Health status values​

Run states​

AI budget alerts​

Next steps​