k8sgpt-ai/k8sgpt-operator

k8sgpt‑operator – AI‑powered analysis & auto‑remediation for Kubernetes

What it is – A Kubernetes Operator that runs the K8sGPT engine inside a cluster. It watches the cluster, sends resource data to a configurable LLM (OpenAI, Azure OpenAI, Bedrock, LocalAI, etc.), receives diagnostic suggestions, and can automatically apply safe, single‑step fixes (e.g., replace a failing container image). All LLM calls are read‑only; the operator itself builds the JSON‑patch and applies it after a policy‑gate and dry‑run.

Key features

• Auto‑remediation (alpha, opt‑in) – Detects failures like ImagePullBackOff, asks the LLM for a fix, generates a semantic patch, dry‑runs it, and rolls out the corrected workload. The LLM never has write permissions.
• Multi‑cluster monitoring – By supplying a kubeconfig secret the operator can manage remote clusters from a central “seed” cluster (useful for Cluster‑API fleets).
• Pluggable AI back‑ends – OpenAI, Azure OpenAI, Amazon Bedrock, LocalAI, and any compatible provider via the spec.ai.backend field.
• Distributed caching – Optional Interplex RPC cache, Azure Blob, or S3 cache to avoid repeated LLM calls.
• Configurable analysis – Interval, resource limits, image pull policy, custom container images, and a rich set of analyzers (core and optional) selectable via spec.filters.
• Result objects – Findings are emitted as Result CRDs, labelled with the originating K8sGPT instance, namespace, and backend for easy filtering.
• Integrations (sinks) – Send results to Slack, Mattermost, CloudEvents, etc.

Installation (Helm)

helm repo add k8sgpt https://charts.k8sgpt.ai/
helm repo update
helm install release k8sgpt/k8sgpt-operator \
  -n k8sgpt-operator-system --create-namespace

Getting started – Create a secret with your LLM credentials, then apply a K8sGPT custom resource. Example for OpenAI:

apiVersion: core.k8sgpt.ai/v1alpha1
kind: K8sGPT
metadata:
  name: k8sgpt-sample
  namespace: k8sgpt-operator-system
spec:
  ai:
    enabled: true
    model: gpt-4o-mini
    backend: openai
    secret:
      name: k8sgpt-sample-secret
      key: openai-api-key
  repository: ghcr.io/k8sgpt-ai/k8sgpt
  version: v0.4.32

The operator will spin up the K8sGPT deployment, start scanning, and create Result objects you can view with kubectl get results.

Enabling auto‑remediation – Add the remediation CR (valid_k8sgpt_remediation_sample.yaml) and set spec.autoRemediation.enabled: true. The operator will then attempt one‑image repairs for detected image‑pull failures, subject to the safety gate described above.

Why it matters – It brings LLM‑driven observability and self‑healing to Kubernetes without giving the LLM direct write access, aligning with security best practices while reducing mean‑time‑to‑resolution for common cluster issues.