This project is a kubectl plugin to generate and apply Kubernetes manifests using OpenAI GPT.

My main motivation is to avoid finding and collecting random manifests when dev/testing things.

Add to brew tap and install with:

brew tap sozercan/kubectl-ai https://github.com/sozercan/kubectl-ai
brew install kubectl-ai

Add to krew index and install with:

kubectl krew index add kubectl-ai https://github.com/sozercan/kubectl-ai
kubectl krew install kubectl-ai/kubectl-ai

• Download the binary from GitHub releases.

• If you want to use this as a kubectl plugin, then copy kubectl-ai binary to your PATH. If not, you can also use the binary standalone.

kubectl-ai requires a valid Kubernetes configuration and one of the following:

• OpenAI API key
• Azure OpenAI Service API key and endpoint
• OpenAI API-compatible endpoint (such as AIKit or LocalAI)

For OpenAI, Azure OpenAI or OpenAI API compatible endpoint, you can use the following environment variables:

export OPENAI_API_KEY=<your OpenAI key>
export OPENAI_DEPLOYMENT_NAME=<your OpenAI deployment/model name. defaults to "gpt-3.5-turbo-0301">
export OPENAI_ENDPOINT=<your OpenAI endpoint, like "https://my-aoi-endpoint.openai.azure.com" or "http://localhost:8080/v1">

If OPENAI_ENDPOINT variable is set, then it will use the endpoint. Otherwise, it will use OpenAI API.

Azure OpenAI service does not allow certain characters, such as ., in the deployment name. Consequently, kubectl-ai will automatically replace gpt-3.5-turbo to gpt-35-turbo for Azure. However, if you use an Azure OpenAI deployment name completely different from the model name, you can set AZURE_OPENAI_MAP environment variable to map the model name to the Azure OpenAI deployment name. For example:

export AZURE_OPENAI_MAP="gpt-3.5-turbo=my-deployment"

• --require-confirmation flag or REQUIRE_CONFIRMATION environment varible can be set to prompt the user for confirmation before applying the manifest. Defaults to true.

• --temperature flag or TEMPERATURE environment variable can be set between 0 and 1. Higher temperature will result in more creative completions. Lower temperature will result in more deterministic completions. Defaults to 0.

• --use-k8s-api flag or USE_K8S_API environment variable can be set to use Kubernetes OpenAPI Spec to generate the manifest. This will result in very accurate completions including CRDs (if present in configured cluster). This setting will use more OpenAI API calls and it requires function calling which is available in 0613 or later models only. Defaults to false. However, this is recommended for accuracy and completeness.

• --k8s-openapi-url flag or K8S_OPENAPI_URL environment variable can be set to use a custom Kubernetes OpenAPI Spec URL. This is only used if --use-k8s-api is set. By default, kubectl-ai will use the configured Kubernetes API Server to get the spec unless this setting is configured. You can use the default Kubernetes OpenAPI Spec or generate a custom spec for completions that includes custom resource definitions (CRDs). You can generate custom OpenAPI Spec by using kubectl get --raw /openapi/v2 > swagger.json.

If you want to use an external editor to edit the generated manifest, you can set the --raw flag and pipe to the editor of your choice. For example:

# Visual Studio Code
$ kubectl ai "create a foo namespace" --raw | code -

# Vim
$ kubectl ai "create a foo namespace" --raw | vim -

$ kubectl ai "create an nginx deployment with 3 replicas"
✨ Attempting to apply the following manifest:
apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment
  labels:
    app: nginx
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: nginx:1.7.9
        ports:
        - containerPort: 80
Use the arrow keys to navigate: ↓ ↑ → ←
? Would you like to apply this? [Reprompt/Apply/Don't Apply]:
+   Reprompt
  ▸ Apply
    Don't Apply

...
Reprompt: update to 5 replicas and port 8080
✨ Attempting to apply the following manifest:
apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deployment
  labels:
    app: nginx
spec:
  replicas: 5
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: nginx:1.7.9
        ports:
        - containerPort: 8080
Use the arrow keys to navigate: ↓ ↑ → ←
? Would you like to apply this? [Reprompt/Apply/Don't Apply]:
+   Reprompt
  ▸ Apply
    Don't Apply

$ kubectl ai "create a foo namespace then create nginx pod in that namespace"
✨ Attempting to apply the following manifest:
apiVersion: v1
kind: Namespace
metadata:
  name: foo
---
apiVersion: v1
kind: Pod
metadata:
  name: nginx
  namespace: foo
spec:
  containers:
  - name: nginx
    image: nginx:latest
Use the arrow keys to navigate: ↓ ↑ → ←
? Would you like to apply this? [Reprompt/Apply/Don't Apply]:
+   Reprompt
  ▸ Apply
    Don't Apply

$ kubectl ai "create a service with type LoadBalancer with selector as 'app:nginx'" --require-confirmation=false
✨ Attempting to apply the following manifest:
apiVersion: v1
kind: Service
metadata:
  name: nginx-service
spec:
  selector:
    app: nginx
  ports:
  - port: 80
    targetPort: 80
  type: LoadBalancer

Please note that the plugin does not know the current state of the cluster (yet?), so it will always generate the full manifest.