Boost Kubernetes Management with AI: Introducing the Lightweight k8m Console

1. Introduction

Kubernetes (K8s) has become ubiquitous, improving development and operations efficiency, yet its complexity poses challenges in networking, troubleshooting, and resource management. With rapid AI advances, intelligent tools can simplify many of these tasks.

2. What is k8m?

k8m is a lightweight console that combines AI and Kubernetes. Built on AMIS and using

kom

as the Kubernetes API client, it embeds the Qwen2.5‑Coder‑7B model and supports private large models such as DeepSeek.

3. Installation and Usage

3.1 ChatGPT Configuration

From version v0.0.8, ChatGPT is built‑in and requires no extra configuration. To use a custom model, set the following environment variables:

<code>export OPENAI_API_KEY="sk-XXXXX"
export OPENAI_API_URL="https://api.siliconflow.cn/v1"
export OPENAI_MODEL="Qwen/Qwen2.5-7B-Instruct"
</code>

Run

./k8m -v 6

for verbose debugging if needed.

3.2 Binary Deployment

Download the latest binary from the GitHub repository https://github.com/weibaohui/k8m/ and start the service:

<code>./k8m
Usage of ./k8m:
  --admin-password string   (default "123456")
  --admin-username string   (default "admin")
  -k, --chatgpt-key string  (default "sk-xxxxxxx")
  -m, --chatgpt-model string (default "Qwen/Qwen2.5-7B-Instruct")
  -u, --chatgpt-url string   (default "https://api.siliconflow.cn/v1")
  -d, --debug               (default false)
  --in-cluster              (default true)
  --jwt-token-secret string (default "your-secret-key")
  -c, --kubeconfig string   (default "/root/.kube/config")
  --kubectl-shell-image string (default "bitnami/kubectl:latest")
  --log-v int               (default 2)
  --login-type string       (default "password")
  --node-shell-image string (default "alpine:latest")
  -p, --port int            (default 3618)
  --sqlite-path string      (default "./data/k8m.db")
  -v, --v Level             (default 2)
</code>

3.3 YAML Deployment in Kubernetes

Key environment variables and their defaults are listed in the table below:

Environment Variable

Default

Description

PORT

3618

Listening port

KUBECONFIG

~/.kube/config

Path to kubeconfig file

OPENAI_API_KEY

""

API key for the large model

OPENAI_API_URL

""

API URL for the large model

OPENAI_MODEL

Qwen/Qwen2.5-7B-Instruct

Default model name (change to deepseek‑ai/DeepSeek‑R1‑Distill‑Qwen‑7B for DeepSeek)

LOGIN_TYPE

"password"

Login method (password, oauth, token, etc.)

ADMIN_USERNAME

"admin"

Administrator username

ADMIN_PASSWORD

"123456"

Administrator password

DEBUG

"false"

Enable debug mode

LOG_V

"2"

klog verbosity level

JWT_TOKEN_SECRET

"your-secret-key"

Secret for JWT token generation

KUBECTL_SHELL_IMAGE

bitnami/kubectl:latest

Image for kubectl shell

NODE_SHELL_IMAGE

alpine:latest

Image for node shell

SQLITE_PATH

/data/k8m.db

SQLite database file path

IN_CLUSTER

"true"

Automatically register host cluster (default enabled)

These variables can be overridden by command‑line arguments.

4. Core Features

4.1 YAML Auto‑Translation

k8m provides integrated YAML browsing, editing, and documentation with automatic translation of YAML fields, eliminating the need to search for attribute meanings.

4.2 Event AI Diagnosis

On the Event page, an AI assistant analyzes abnormal events and provides detailed explanations with a single click.

4.3 Log AI Diagnosis

Select error logs in the Pod log view and click the AI diagnosis button to receive a diagnostic report that pinpoints key errors and suggests solutions.

4.4 AI‑Driven Inspection

The AI inspection feature leverages predefined rules from the k8s_gpt tool to automatically audit clusters, nodes, deployments, stateful sets, services, ingress, and other resources.

4.5 Command Generation

Describe the desired operation in natural language, and k8m generates the appropriate shell command, reducing lookup time.

4.6 Multi‑Cluster Management (MCP)

Version 0.0.66 adds support for managing multiple clusters from a single dashboard via kubeconfig files.

4.7 Role‑Based Access Control

k8m supports adding users and groups, assigning three role types: Cluster Administrator (full control), Cluster Read‑Only (view only), and Platform Administrator (manage platform settings).

5. Conclusion

k8m combines AI with Kubernetes to deliver a concise, efficient console that streamlines file management, log diagnostics, YAML translation, and fault isolation. Its intuitive UI and AI capabilities greatly improve Kubernetes usability for developers and operators, while supporting private model deployment for enhanced data security and autonomy.