Incorrect response with Local AI and k8sgpt

[dilipk8]

LocalAI version:
quay.io/go-skynet/local-ai:latest

Environment, CPU architecture, OS, and Version:
5.13.0-1022-aws #24~20.04.1-Ubuntu SMP x86_64 x86_64 x86_64 GNU/Linux
Kubernetes version - 1.22.4

Describe the bug
I am trying to use k8sgpt with Local AI. My K8s cluster is a kops cluster on AWS. I configured local AI as per the docs using the ggml-gpt4all-j model.
I create an nginx pod with an invalid image and this was the k8sgpt analyze result, which is incorrect.

• Error: Back-off pulling image "nginedd"
  The prompt is asking to simulate a Kubernet error message in English, delimited by triple dashes. The user is then asked to provide a step-by-step solution, in no more than 280 characters. The output should be in the format: Error (explain error here) Solution (step-by-step solution here).

To Reproduce
git clone https://github.com/go-skynet/LocalAI
cd LocalAI
wget https://gpt4all.io/models/ggml-gpt4all-j.bin -O models/ggml-gpt4all-j
cp -rf prompt-templates/ggml-gpt4all-j.tmpl models/
docker-compose up -d --pull always

Next - Use a K8s cluster, install and configure k8sgpt cli with local AI and run k8sgpt analyze and see the result.

Expected behavior
Valid message which tell us how to troubleshoot the invalid image error.

[mudler]

seems more a limitation on the model. try for Koala for instance https://github.com/go-skynet/model-gallery#gpt-koala

[dilipk8]

@mudler , okay,but just wondering it looks like the same model worked fine, in your video here - https://www.youtube.com/watch?v=PKrDNuJ_dfE
I will try another model but is there something am missing out or something that i can double-check

[vitorstone]

So you installed LocalAI on a container outside of the cluster, and then configured the k8sgpt operator on the cluster. How did you configure it so that it works with the LocalAI API outside the cluster?