Configure telemetry on policies to export metrics and traces from the Actions API. You can enable request-level and plugin-level traces, attach extra metadata, and send data to multiple exporters such as Kafka and TrustLens.

​
Telemetry settings

  • enable_request_traces (bool): Export traces at the request level
  • enable_plugin_traces (bool): Export traces at the plugin/action level
  • extra_params (object): Key-value metadata to include with telemetry
  • exporters (array): List of exporters with name and provider-specific settings
  • header_mapping (object): Map internal telemetry fields to incoming HTTP headers

​
Example: policy with telemetry

{
  "name": "policy-with-telemetry",
  "actions": [
    {"name": "sanitize", "priority": 1, "parallel": false, "settings": {"level": "standard"}},
    {"name": "classify", "priority": 2, "parallel": false, "settings": {"labels": ["spam", "ham"]}}
  ],
  "telemetry": {
    "enable_request_traces": true,
    "enable_plugin_traces": true,
    "extra_params": {
      "app_id": "12345"
    },
    "exporters": [
      {
        "name": "kafka",
        "settings": {
          "host": "localhost",
          "port": "9092",
          "topic": "metrics"
        }
      },
      {
        "name": "trustlens",
        "settings": {
          "host": "localhost",
          "mapping": {
            "input": {
              "data_projection": {
                "input": "payload.input"
              }
            },
            "output": {}
          },
          "port": "9092",
          "topic": "traces"
        }
      }
    ],
    "header_mapping": {
      "conversation_id": "X-CONV-ID",
      "interaction_id": "X-INT-ID",
      "session_id": "X-SESSION-ID"
    }
  }
}

​
Tips

  • Start with request traces, then add plugin traces for deeper diagnostics
  • Use header mapping to propagate correlation IDs from your clients
  • Include stable fields in extra_params (e.g., app_id, tenant) to segment telemetry
  • Send data to multiple exporters if you use more than one backend (e.g., Kafka + TrustLens)