i'm trying to figure out how to instantiate a kubernetes provider from a GCP Cluster object that we've also created in Pulumi, without needing to generate my own kubeconfig out-of-band. this answer references an output in the typescript API,

cluster.kubeConfigRaw

, that does not appear in the docs. if you ask the AI to rewrite the program in Python (which we use), it references a

kube_config

output that is also not documented, and, afaict, doesn't actually exist:

AttributeError: 'Cluster' object has no attribute 'kube_config'

so it seems the AI is hallucinating. is this possible? can i take outputs from the

gcp.container.Cluster

object and directly instantiate a k8s provider?

The Pulumi AI has been both helpful and extremely frustrating for my team to use. Lots and lots of hallucinations and code that is clearly broken once copied/pasted into the editor. 😕

There's an example here of how to create the kubeconfig: https://github.com/pulumi/examples/blob/00393ba60105dee7de3ac43909c07dc0573c2374/gcp-py-gke/__main__.py#L71

1. that's helpful, thanks! 2. wow that's ugly 😬

Yep. Personally, I make it with a normal dict, then yaml encode it

that's definitely less ugly!

Feel free to upvote this issue: https://github.com/pulumi/pulumi-gcp/issues/1064