Docs
Documentation
Agentgateway
LLM consumption
View traces

View traces

Use a custom ConfigMap to configure your agentgateway proxy for tracing.

Before you begin

  1. Install kgateway and enable the agentgateway integration.

  2. Verify that the agentgateway integration is enabled.

    helm get values kgateway -n kgateway-system -o yaml

    Example output:

    agentgateway:
  enabled: true

Set up an OpenTelemetry collector

Install an OpenTelemetry collector that the agentgateway proxy can send traces to. Depending on your environment, you can further configure your OpenTelemetry to export these traces to your preferred tracing platform, such as Jaeger.

  1. Install the OTel collector.

    helm upgrade --install opentelemetry-collector-traces opentelemetry-collector \
--repo https://open-telemetry.github.io/opentelemetry-helm-charts \
--version 0.127.2 \
--set mode=deployment \
--set image.repository="otel/opentelemetry-collector-contrib" \
--set command.name="otelcol-contrib" \
--namespace=telemetry \
--create-namespace \
-f -<<EOF
config:
  receivers:
    otlp:
      protocols:
        grpc:
          endpoint: 0.0.0.0:4317
        http:
          endpoint: 0.0.0.0:4318
  exporters:
    otlp/tempo:
      endpoint: http://tempo.telemetry.svc.cluster.local:4317
      tls:
        insecure: true
    debug:
      verbosity: detailed
  service:
    pipelines:
      traces:
        receivers: [otlp]
        processors: [batch]
        exporters: [debug, otlp/tempo]
EOF

  2. Verify that the collector is up and running.

    kubectl get pods -n telemetry

    Example output:

    NAME                                             READY   STATUS    RESTARTS   AGE
opentelemetry-collector-traces-8f566f445-l82s6   1/1     Running   0          17m

Configure your proxy

  1. Create a ConfigMap with your agentgateway tracing configuration. The following example collects additional information about the request to the LLM and adds this information to the trace. The trace is then sent to the collector that you set up earlier. To learn more about the fields that you can configure, see the agentgateway docs.

    ℹ️
    For more tracing providers, see Other tracing configurations.
    kubectl apply -f- <<EOF
apiVersion: v1
kind: ConfigMap
metadata:
  name: agent-gateway-config
  namespace: kgateway-system
data:
  config.yaml: |-
    config:
      tracing:
        otlpEndpoint: http://opentelemetry-collector-traces.telemetry.svc.cluster.local:4317
        otlpProtocol: grpc
        randomSampling: true
        fields:
          add:
            gen_ai.operation.name: '"chat"'
            gen_ai.system: "llm.provider"
            gen_ai.request.model: "llm.requestModel"
            gen_ai.response.model: "llm.responseModel"
            gen_ai.usage.completion_tokens: "llm.outputTokens"
            gen_ai.usage.prompt_tokens: "llm.inputTokens"    
EOF

  2. Create a GatewayParameters resource that references the ConfigMap that you created.

    kubectl apply -f- <<EOF
apiVersion: gateway.kgateway.dev/v1alpha1
kind: GatewayParameters
metadata:
  name: tracing
  namespace: kgateway-system
spec:
  kube:
    agentgateway:
      customConfigMapName: agent-gateway-config
EOF

  3. Create your agentgateway proxy. Make sure to reference the GatewayParameters resource that you created so that your proxy starts with the custom tracing configuration.

    kubectl apply -f- <<EOF
kind: Gateway
apiVersion: gateway.networking.k8s.io/v1
metadata:
  name: agentgateway
  namespace: kgateway-system
  labels:
    app: agentgateway
spec:
  gatewayClassName: agentgateway
  infrastructure:
    parametersRef:
      name: tracing
      group: gateway.kgateway.dev  
      kind: GatewayParameters  
  listeners:
  - protocol: HTTP
    port: 8080
    name: http
    allowedRoutes:
      namespaces:
        from: All
EOF

  4. Verify that your agentgateway proxy is up and running.

    kubectl get pods -n kgateway-system

    Example output:

    NAMESPACE            NAME                                         READY   STATUS    RESTARTS   AGE
kgateway-system      agentgateway-8b5dc4874-bl79q         1/1     Running   0          12s

  5. Get the external address of the gateway and save it in an environment variable. 

    export INGRESS_GW_ADDRESS=$(kubectl get svc -n kgateway-system agentgateway -o jsonpath="{.status.loadBalancer.ingress[0]['hostname','ip']}")
echo $INGRESS_GW_ADDRESS  
    kubectl port-forward deployment/agentgateway -n kgateway-system 8080:8080

Set up access to Gemini

Configure access to an LLM provider such as Gemini and send a sample request. You later use this request to verify your tracing configuration.

  1. Save your Gemini API key as an environment variable. To retrieve your API key, log in to the Google AI Studio and select API Keys.

    export GOOGLE_KEY=<your-api-key>

  2. Create a secret to authenticate to Google.

    kubectl apply -f - <<EOF
apiVersion: v1
kind: Secret
metadata:
  name: google-secret
  namespace: kgateway-system
  labels:
    app: agentgateway
type: Opaque
stringData:
  Authorization: $GOOGLE_KEY
EOF

  3. Create a Backend resource to define the Gemini destination.

    kubectl apply -f- <<EOF
apiVersion: gateway.kgateway.dev/v1alpha1
kind: Backend
metadata:
  labels:
    app: agentgateway
  name: google
  namespace: kgateway-system
spec:
  ai:
    llm:
     gemini:
          apiVersion: v1beta
          authToken:
            kind: SecretRef
            secretRef:
              name: google-secret
          model: gemini-1.5-flash-latest
  type: AI
EOF
    Review the following table to understand this configuration.
    Setting Description
    gemini The Gemini AI provider.
    apiVersion The API version of Gemini that is compatible with the model that you plan to use. In this example, you must use v1beta because the gemini-1.5-flash-latest model is not compatible with the v1 API version. For more information, see the Google AI docs.
    authToken The authentication token to use to authenticate to the LLM provider. The example refers to the secret that you created in the previous step.
    model The model to use to generate responses. In this example, you use the gemini-1.5-flash-latest model. For more models, see the Google AI docs.

  4. Create an HTTPRoute resource to route requests to the Gemini backend. Note that kgateway automatically rewrites the endpoint that you set up (such as /gemini) to the appropriate chat completion endpoint of the LLM provider for you, based on the LLM provider that you set up in the Backend resource.

    kubectl apply -f- <<EOF
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: google
  namespace: kgateway-system
  labels:
    app: agentgateway
spec:
  parentRefs:
    - name: agentgateway
      namespace: kgateway-system
  rules:
  - matches:
    - path:
        type: PathPrefix
        value: /gemini
    backendRefs:
    - name: google
      namespace: kgateway-system
      group: gateway.kgateway.dev
      kind: Backend
EOF

  5. Send a request to the LLM provider API. Verify that the request succeeds and that you get back a response from the chat completion API.

    curl -vik "$INGRESS_GW_ADDRESS:8080/gemini" -H content-type:application/json  -d '{
  "model": "",
  "messages": [
   {"role": "user", "content": "Explain how AI works in simple terms."}
 ]
}'
    curl -vik "localhost:8080/gemini" -H content-type:application/json  -d '{
  "model": "",
  "messages": [
   {"role": "user", "content": "Explain how AI works in simple terms."}
 ]
}'

    Example output:

    {"id":"aGLEaMjbLp6p_uMPopeAoAc",
"choices":
  [{"index":0,"message":{
      "content":"Imagine teaching a dog a trick.  You show it what to do, reward it when it's right, and correct it when it's wrong.  Eventually, the dog learns.\n\nAI is similar.  We \"teach\" computers by showing them lots of examples.  For example, to recognize cats in pictures, we show it thousands of pictures of cats, labeling each one \"cat.\"  The AI learns patterns in these pictures – things like pointy ears, whiskers, and furry bodies – and eventually, it can identify a cat in a new picture it's never seen before.\n\nThis learning process uses math and algorithms (like a secret code of instructions) to find patterns and make predictions.  Some AI is more like a dog learning tricks (learning from examples), and some is more like following a very detailed recipe (following pre-programmed rules).\n\nSo, in short: AI is about teaching computers to learn from data and make decisions or predictions, just like we teach dogs tricks.\n",
      "role":"assistant"
      },
   "finish_reason":"stop"
   }],
 "created":1757700714,
 "model":"gemini-1.5-flash-latest",
 "object":"chat.completion",
 "usage":{
     "prompt_tokens":8,
     "completion_tokens":205,
     "total_tokens":213
     }
}

Verify tracing

  1. Get the logs of the agentgateway proxy. In the CLI output, find the trace.id.

    kubectl logs deploy/agentgateway -n kgateway-system

    Example output:

    info	request gateway=kgateway-system/agentgateway listener=http 
route=kgateway-system/google endpoint=generativelanguage.googleapis.com:443
src.addr=127.0.0.1:49576 http.method=POST http.host=localhost 
http.path=/gemini http.version=HTTP/1.1 http.status=200 
trace.id=d65e4eeb983e2d964e71e8dc8c405f97 span.id=b836e1b1d51b3e74 
llm.provider=gemini llm.request.model= llm.request.tokens=8 
llm.response.model=gemini-1.5-flash-latest llm.response.tokens=313 duration=3165ms

  2. Get the logs of the collector and search for the trace ID. Verify that you see the additional LLM that you configured initially.

    kubectl logs deploy/opentelemetry-collector-traces -n telemetry

    Example output:

    Span #0
 Trace ID       : d65e4eeb983e2d964e71e8dc8c405f97
 Parent ID      : 
 ID             : b836e1b1d51b3e74
 Name           : POST /gemini
 Kind           : Server
 Start time     : 2025-09-24 18:12:58.653868462 +0000 UTC
 End time       : 2025-09-24 18:13:01.821700755 +0000 UTC
 Status code    : Unset
 Status message : 
Attributes:
  -> gateway: Str(kgateway-system/agentgateway)
  -> listener: Str(http)
  -> route: Str(kgateway-system/google)
  -> endpoint: Str(generativelanguage.googleapis.com:443)
  -> src.addr: Str(127.0.0.1:49576)
  -> http.method: Str(POST)
  -> http.host: Str(localhost)
  -> http.path: Str(/gemini)
  -> http.version: Str(HTTP/1.1)
  -> http.status: Int(200)
  -> trace.id: Str(d65e4eeb983e2d964e71e8dc8c405f97)
  -> span.id: Str(b836e1b1d51b3e74)
  -> llm.provider: Str(gemini)
  -> llm.request.model: Str()
  -> llm.request.tokens: Int(8)
  -> llm.response.model: Str(gemini-1.5-flash-latest)
  -> llm.response.tokens: Int(313)
  -> duration: Str(3165ms)
  -> url.scheme: Str(http)
  -> network.protocol.version: Str(1.1)
  -> gen_ai.operation.name: Str(chat)
  -> gen_ai.system: Str(gemini)

Other tracing configurations

Review common tracing providers configurations that you can use with agentgateway.

apiVersion: v1
kind: ConfigMap
metadata:
  name: jaeger-tracing-config
data:
  config.yaml: |-
    config:
      tracing:
        otlpEndpoint: http://jaeger-collector.jaeger.svc.cluster.local:4317
        otlpProtocol: grpc
        randomSampling: true
        fields:
          add:
            gen_ai.operation.name: '"chat"'
            gen_ai.system: "llm.provider"
            gen_ai.request.model: "llm.request_model"
            gen_ai.response.model: "llm.response_model"
            gen_ai.usage.completion_tokens: "llm.output_tokens"
            gen_ai.usage.prompt_tokens: "llm.input_tokens"    
apiVersion: v1
kind: ConfigMap
metadata:
  name: langfuse-tracing-config
data:
  config.yaml: |-
    config:
      tracing:
        otlpEndpoint: https://us.cloud.langfuse.com/api/public/otel
        otlpProtocol: http
        headers:
          Authorization: "Basic <base64-encoded-credentials>"
        randomSampling: true
        fields:
          add:
            gen_ai.operation.name: '"chat"'
            gen_ai.system: "llm.provider"
            gen_ai.prompt: "llm.prompt"
            gen_ai.completion: 'llm.completion.map(c, {"role":"assistant", "content": c})'
            gen_ai.usage.completion_tokens: "llm.output_tokens"
            gen_ai.usage.prompt_tokens: "llm.input_tokens"
            gen_ai.request.model: "llm.request_model"
            gen_ai.response.model: "llm.response_model"
            gen_ai.request: "flatten(llm.params)"    
apiVersion: v1
kind: ConfigMap
metadata:
  name: phoenix-tracing-config
data:
  config.yaml: |-
    config:
      tracing:
        otlpEndpoint: http://localhost:4317
        randomSampling: true
        fields:
          add:
            span.name: '"openai.chat"'
            openinference.span.kind: '"LLM"'
            llm.system: "llm.provider"
            llm.input_messages: 'flatten_recursive(llm.prompt.map(c, {"message": c}))'
            llm.output_messages: 'flatten_recursive(llm.completion.map(c, {"role":"assistant", "content": c}))'
            llm.token_count.completion: "llm.output_tokens"
            llm.token_count.prompt: "llm.input_tokens"
            llm.token_count.total: "llm.total_tokens"    
apiVersion: v1
kind: ConfigMap
metadata:
  name: openllmetry-tracing-config
data:
  config.yaml: |-
    config:
      tracing:
        otlpEndpoint: http://localhost:4317
        randomSampling: true
        fields:
          add:
            span.name: '"openai.chat"'
            gen_ai.operation.name: '"chat"'
            gen_ai.system: "llm.provider"
            gen_ai.prompt: "flatten_recursive(llm.prompt)"
            gen_ai.completion: 'flatten_recursive(llm.completion.map(c, {"role":"assistant", "content": c}))'
            gen_ai.usage.completion_tokens: "llm.output_tokens"
            gen_ai.usage.prompt_tokens: "llm.input_tokens"
            gen_ai.request.model: "llm.request_model"
            gen_ai.response.model: "llm.response_model"
            gen_ai.request: "flatten(llm.params)"
            llm.is_streaming: "llm.streaming"

Cleanup

You can remove the resources that you created in this guide. 
kubectl delete gateway agentgateway -n kgateway-system
kubectl delete GatewayParameters   tracing -n kgateway-system
kubectl delete configmap agent-gateway-config -n kgateway-system
helm uninstall opentelemetry-collector-traces -n telemetry
kubectl delete httproute google -n kgateway-system
kubectl delete backend google -n kgateway-system
kubectl delete secret google-secret -n kgateway-system
View metrics and logs