Skip to content

K8sGPT Operator


  • To begin you will require a Kubernetes cluster available and KUBECONFIG set.
  • You will also need to install helm v3. See the Helm documentation for more information.

Operator Installation

To install the operator, run the following command:

helm repo add k8sgpt
helm repo update
helm install release k8sgpt/k8sgpt-operator -n k8sgpt-operator-system --create-namespace

This will install the Operator into the cluster, which will await a K8sGPT resource before anything happens.

Deploying an OpenAI secret

Whilst there are multiple backends supported ( OpenAI, Azure OpenAI and Local ), in this example we'll use OpenAI. Whatever backend you are using, you need to make sure to have a secret that works with the backend.

For instance, this means you will need to install your OpenAI token as a secret into the cluster:

kubectl create secret generic k8sgpt-sample-secret --from-literal=openai-api-key=$OPENAI_TOKEN -n k8sgpt-operator-system

Deploying a K8sGPT resource

To deploy a K8sGPT resource, you will need to create a YAML file with the following contents:

kubectl apply -f - << EOF
kind: K8sGPT
  name: k8sgpt-sample
  namespace: k8sgpt-operator-system
    enabled: true
    model: gpt-3.5-turbo
    backend: openai
      name: k8sgpt-sample-secret
      key: openai-api-key
    # anonymized: false
    # language: english
  noCache: false
  version: v0.3.17
  # filters:
  #   - Ingress
  # sink:
  #   type: slack
  #   webhook: <webhook-url>
  # extraOptions:
  #   backstage:
  #     enabled: true

Please replace the <VERSION> field with the current release of K8sGPT. At the time of writing this is v0.3.17.

Regarding out-of-cluster traffic to AI backends

In the above example enableAI is set to true. This option allows the cluster deployment to use the backend to filter and improve the responses to the user. Those responses will appear as details within the Result custom resources that are created.

The default backend in this example is OpenAI and allows for additional details to be generated and solutions provided for issues. If you wish to disable out-of-cluster communication and any Artificial Intelligence processing through models, simply set enableAI to false.

It should also be noted that localai and azureopenai are supported and in-cluster models will be supported in the near future

Viewing the results

Once the initial scans have been completed after several minutes, you will be presented with results custom resources.

❯ kubectl get results -o json | jq .
  "apiVersion": "v1",
  "items": [
      "apiVersion": "",
      "kind": "Result",
      "metadata": {
        "creationTimestamp": "2023-04-26T09:45:02Z",
        "generation": 1,
        "name": "placementoperatorsystemplacementoperatorcontrollermanagermetricsservice",
        "namespace": "default",
        "resourceVersion": "108371",
        "uid": "f0edd4de-92b6-4de2-ac86-5bb2b2da9736"
      "spec": {
        "details": "The error message means that the service in Kubernetes doesn't have any associated endpoints, which should have been labeled with \"control-plane=controller-manager\". \n\nTo solve this issue, you need to add the \"control-plane=controller-manager\" label to the endpoint that matches the service. Once the endpoint is labeled correctly, Kubernetes can associate it with the service, and the error should be resolved.",