[Question]: failed while calling AI provider localai: error, status code: 401 about k8sgpt-operator HOT 9 CLOSED

hey @wmashal the README file looks wrong for localai, you have to specify the baseURL and the local model you 've deployed; use also the latest operator version with v0.3.4 version

also a post related to localai with the operator (might be slightly outdated)
the 401s are generated on the absence of a baseURL it will fetch the default openAIs endpoint thus complaining for a password.

fyi, disabling the enableAI flag will just produce the errors from the analysers without any explanation from your local model

  ...
  model: ggml-gpt4all-j.bin
  # kubernetes-internal DNS name of the local-ai Service
  baseUrl: http://local-ai.local-ai.svc.cluster.local:8080/v1
  ...

from k8sgpt-operator.

additional note I can call the loacalAI directly

from k8sgpt-operator.

I am not sure if this is the result but when I change enableAI to false I start see results , please advice ? also why I cant see suggested solution in the result CR ?

from k8sgpt-operator.

Hi @arbreezy
Thanks for you reply

but I am still facing the issue, as you recommended I am using v0.3.4 and the enableAI is true but still see this error in the k8gpt deployment

As you see I have the baseURL but also the operator still complaining

from k8sgpt-operator.

Can you call that endpoint (local-ai.local-ai.svc.cluster.local:8080) from a busybox container with curl?

from k8sgpt-operator.

@wmashal based on this screenshot you have installed LocalAI in the k8sgpt-local-ai namespace, therefore your baseUrl should be: http://local-ai.k8sgpt-local-ai.svc.cluster.local:8080/v1. The format for baseUrl is <service_name>.<namespace>.svc.cluster.local:8080/v1, where you plug in the service name and namespace where LocalAI is deployed.

from k8sgpt-operator.

@TylerGillson

Thanks for your reply
Actually the namespace not the issue, I create a clean new kind cluster then I deployed the operator, localAI and the k8sGPT in the default namespace and the baseURL also point to default, but still same issue, see the attached images

from k8sgpt-operator.

@wmashal there are two separate issues present in your latest screenshot.

The first issue is that the operator is reconciling two K8sGPT resources, one in the default namespace and one in the local-ai namespace, which doesn't exist. Please delete the K8sGPT that's using the local-ai namespace.

I am uncertain about the second issue: error reading server preface: http2: frame too large. Perhaps @matthisholleville could comment.

from k8sgpt-operator.

Actually, @wmashal, the version field in the latest K8sGPT resource you shared is very old. Please delete and recreate your K8sGPT using version v0.3.4 and try again.

from k8sgpt-operator.