This is a sample code: just to illustrate the methodology of do k8s(kubectl) operation with LLM assistant.

Thanks to project Kubectl-GPT for the great prompt as my reference.

Usage:

pip instal openai

export OPENAI_API_KEY="$your-key"
python3 kubectl-chat.py "List pods which is Pending and show their pending reason"

If using openAI proxy, export environment variables before run the script:

export OPENAI_API_BASE=http://$your-proxy
export OPENAI_API_KEY="$your-key"
export MODEL="$other-model-name" #"gpt-3.5-turbo-16k"

Theory:

1. Create a prompt to ask LLM(openAI by default) to acts as a script generator
2. Show LLM some example (good example and bad ones)
3. execute the shell script locally

Relevant projects:

• https://github.com/feiskyer/kube-copilot
• https://github.com/knight42/kopilot
• https://github.com/sozercan/kubectl-ai
• https://github.com/abhishek-ch/Kubectl-GPT