Ollama Gemma4 support

Author: skb888Created Apr 14, 2026Updated Apr 18, 2026

Hi, I have tried kubectl-ai:v0.0.31 and self-hosted gemma4:latest from ollama. I have used below configure: command: [ "--llm-provider", "ollama", "--model", "gemma4:latest", "--enable-tool-use-shim", "--ui-listen-address", "0.0.0.0:8088", "--ui-type", "web", "--kubeconfig", "/etc/kubernetes/read-only-kubeconfig.conf" ] While, It seems that --enable-tool-use-shim not work for gemma4 since I faced Error: parsing ReAct response "Hello! I'm kubectl-ai, your AI assistant specializing in Kubernetes operations.\n\nHow can I help you with your cluster today? Are you looking to deploy an application, troubleshoot a failing pod, check cluster status, or something else? Just let me know what you need! ✨": no JSON code block found in "". Gemma4 has special token protocol. Can you suggest what I can do? Thanks a lot.

Source: GoogleCloudPlatform/kubectl-ai