Skip to main content

Configure LLM Providers for KinD (Optional)

This page is optional. If you used Run CAIPE with KinD and ran ./setup-caipe.sh, the script already prompted you for an LLM provider and API key and stored them in the cluster. You can skip this page unless you need to:

  • Change the LLM provider or model after setup
  • Add or rotate API keys without re-running the full setup
  • Configure LLM access for automation (e.g. non-interactive or CI)

For the standard flow, the KinD setup guide (Step 3) is enough.


When you need this

Use the steps below when you want to update LLM configuration manually (e.g. via kubectl or Helm) instead of re-running ./setup-caipe.sh. The exact secret names and keys depend on your deployment; they are typically created by the setup script in the ai-platform-engineering (or similar) namespace.

  • Anthropic Claude: Store the API key in a Kubernetes secret; the setup script can also prompt for it interactively.
  • OpenAI: Same idea—use a Kubernetes secret or re-run the setup script and choose OpenAI when prompted.
  • AWS Bedrock: Use AWS credentials (env vars or ~/.aws/credentials) or a Kubernetes secret as described in the KinD setup section.

For full options and environment variables, see Run CAIPE with KinD.