Destinations

Configure where the OpenLIT Operator sends AI telemetry — traces, metrics, and logs — by setting the OTLP endpoint and optional authentication headers in your AutoInstrumentation resource.

The OpenLIT Operator forwards telemetry to any OTLP-compatible backend. You configure the destination in the spec.otlp field of your AutoInstrumentation resource. This lets you consolidate AI observability data in your existing monitoring stack without additional pipeline components.

How destination routing works

Every AutoInstrumentation resource has its own spec.otlp configuration. This means you can route different services to different backends by creating multiple AutoInstrumentation resources with different selectors and endpoints.

spec:
  otlp:
    endpoint: "https://your-backend:4318"   # Required — OTLP HTTP endpoint
    headers: "authorization=Bearer token"    # Optional — authentication headers
    timeout: 30                              # Optional — seconds, default 30

OpenLIT Platform

Send telemetry to the OpenLIT observability platform deployed in your cluster. This is the recommended setup when running OpenLIT alongside the operator.

apiVersion: openlit.io/v1alpha1
kind: AutoInstrumentation
metadata:
  name: openlit-platform-instrumentation
  namespace: default
spec:
  selector:
    matchLabels:
      instrument: "true"
  python:
    instrumentation:
      provider: "openlit"
      version: "latest"
  otlp:
    endpoint: "http://openlit.openlit.svc.cluster.local:4318"
    timeout: 30
  resource:
    environment: "production"
    serviceName: "my-ai-service"

When the OpenLIT Platform is deployed in the same cluster, the endpoint http://openlit.openlit.svc.cluster.local:4318 is the fully qualified in-cluster service address. If OpenLIT is deployed externally, use your external endpoint instead.

Grafana Cloud

Send telemetry to Grafana Cloud’s OTLP gateway. Retrieve your gateway URL and authentication header from Grafana Cloud Portal → OpenTelemetry.

apiVersion: openlit.io/v1alpha1
kind: AutoInstrumentation
metadata:
  name: grafana-cloud-instrumentation
  namespace: default
spec:
  selector:
    matchLabels:
      instrument: "true"
  python:
    instrumentation:
      provider: "openlit"
      version: "latest"
  otlp:
    endpoint: "https://otlp-gateway-<zone>.grafana.net/otlp"
    headers: "Authorization=Basic%20<base64-encoded-instance-id-and-api-token>"
    timeout: 30
  resource:
    environment: "production"
    serviceName: "my-ai-service"

Replace:

• https://otlp-gateway-<zone>.grafana.net/otlp with the OTEL_EXPORTER_OTLP_ENDPOINT value from the Grafana Cloud portal
• Authorization=Basic%20... with the OTEL_EXPORTER_OTLP_HEADERS value from the portal

Prometheus + Jaeger (via OpenTelemetry Collector)

To route metrics to Prometheus and traces to Jaeger, deploy an OpenTelemetry Collector configured with both exporters and point the operator at the collector’s OTLP endpoint.

apiVersion: openlit.io/v1alpha1
kind: AutoInstrumentation
metadata:
  name: otelcol-instrumentation
  namespace: default
spec:
  selector:
    matchLabels:
      instrument: "true"
  python:
    instrumentation:
      provider: "openlit"
      version: "latest"
  otlp:
    endpoint: "http://otel-collector.monitoring.svc.cluster.local:4318"
    timeout: 30
  resource:
    environment: "production"
    serviceName: "my-ai-service"

Ensure your OpenTelemetry Collector is configured to export metrics to Prometheus and traces to Jaeger before routing telemetry to it.

Datadog

Send telemetry to Datadog through the Datadog Agent. The Datadog Agent handles authentication and forwards data to the Datadog platform.

apiVersion: openlit.io/v1alpha1
kind: AutoInstrumentation
metadata:
  name: datadog-instrumentation
  namespace: default
spec:
  selector:
    matchLabels:
      instrument: "true"
  python:
    instrumentation:
      provider: "openlit"
      version: "latest"
  otlp:
    endpoint: "http://datadog-agent.default.svc.cluster.local:4318"
    timeout: 30
  resource:
    environment: "production"
    serviceName: "my-ai-service"

Replace datadog-agent.default.svc.cluster.local with your actual Datadog Agent service name and namespace. The agent must be configured to accept OTLP traffic.

SigNoz

Send telemetry directly to your SigNoz instance using an ingestion key for authentication.

apiVersion: openlit.io/v1alpha1
kind: AutoInstrumentation
metadata:
  name: signoz-instrumentation
  namespace: default
spec:
  selector:
    matchLabels:
      instrument: "true"
  python:
    instrumentation:
      provider: "openlit"
      version: "latest"
  otlp:
    endpoint: "https://ingest.{region}.signoz.cloud:443"
    headers: "signoz-access-token=YOUR_SIGNOZ_INGESTION_KEY"
    timeout: 30
  resource:
    environment: "production"
    serviceName: "my-ai-service"

Replace:

• https://ingest.{region}.signoz.cloud:443 with your SigNoz endpoint URL
• YOUR_SIGNOZ_INGESTION_KEY with your SigNoz access token (e.g. 896cc137-xxxx-xxxx-b8a0-a6d7ab65f274)

New Relic

Send telemetry directly to New Relic’s OTLP endpoint using your New Relic license key.

apiVersion: openlit.io/v1alpha1
kind: AutoInstrumentation
metadata:
  name: newrelic-instrumentation
  namespace: default
spec:
  selector:
    matchLabels:
      instrument: "true"
  python:
    instrumentation:
      provider: "openlit"
      version: "latest"
  otlp:
    endpoint: "https://otlp.nr-data.net:443"
    headers: "api-key=YOUR_NEWRELIC_LICENSE_KEY"
    timeout: 30
  resource:
    environment: "production"
    serviceName: "my-ai-service"

Replace YOUR_NEWRELIC_LICENSE_KEY with your New Relic license key (e.g. acxxxxxxxxxxxxxxxx).

Authentication patterns

The spec.otlp.headers field accepts comma-separated key=value pairs. Use it to pass any HTTP headers required by your backend.

Backend type	Header format	Example
Bearer token	authorization=Bearer <token>	authorization=Bearer eyJ...
API key	x-api-key=<key>	x-api-key=secret456
SigNoz	signoz-access-token=<key>	signoz-access-token=896cc...
New Relic	api-key=<license-key>	api-key=acxxxxxxxxxx
Multiple headers	key1=val1,key2=val2	authorization=Bearer t,x-tenant=prod

Store sensitive values like API keys in Kubernetes Secrets and reference them via spec.python.instrumentation.env[].valueFrom.secretKeyRef rather than hardcoding them in the AutoInstrumentation resource.