DOCSEARCHbyAlgolia
Submit Docs

Headlamp AI Assistant

repository·main·Indexed 41 days ago

https://github.com/headlamp-k8s/plugins

An alpha-state plugin for the Headlamp Kubernetes dashboard providing a conversational interface for cluster management. Key features include natural language resource generation, context-aware troubleshooting, and support for Model Context Protocol (MCP) servers in the desktop application. It includes a companion CLI (@headlamp-k8s/ai-cli) for one-shot queries and interactive REPL sessions with support for multiple AI providers including OpenAI, Anthropic, Gemini, and GitHub Copilot.

Tokens
172.7K
Snippets
427
Records
687
Agent score
54%

What's inside Headlamp AI Assistant

  1. Overview of the Kubeflow Training Plugin

    main

    The Kubeflow Training Plugin provides platform operators, SREs, and ML engineers with Kubernetes-native observability for Kubeflow Training workloads using the unified Training Operator v2 architecture.

    Key capabilities include:

    • Hierarchical Resource Access: Direct visibility into JobSet, Job, and Pod resources created by a TrainJob.
    • Resource Monitoring: Immediate visibility into GPU, CPU, and Memory usage for training nodes.
    • Runtime Inspection: Visual diffs and JSON views of how TrainingRuntime or ClusterTrainingRuntime are patched into TrainJob resources.
    • Scheduling Insights: Visibility into Volcano or Coscheduling configurations for distributed workloads.

  2. Overview of the Kueue Headlamp Plugin

    main

    The Kueue plugin integrates Kubernetes batch workload queueing and admission visibility into the Headlamp UI. It provides dedicated views for Kueue resources, allowing users to inspect queue configurations, workload admission, resource usage, and the relationships between different Kueue components without using kubectl or raw CRDs.

    This plugin supports the following kueue.x-k8s.io/v1beta2 resources:

    • Cohort
    • ClusterQueue
    • LocalQueue
    • ResourceFlavor
    • Workload

  3. Overview of Headlamp Plugins

    main

    This repository contains official plugins for Headlamp, a Kubernetes UI. Plugins extend Headlamp's functionality by adding support for specific Kubernetes resources, tools, and integrations (e.g., AI assistants, Helm chart management, or observability tools like Prometheus).

    Plugins are categorized into:

    1. Current Plugins: Official plugins maintained within this repository.
    2. External Plugins: Community-maintained plugins often published via Artifact Hub.

    Developers looking to learn how to build plugins should refer to the examples folder in the main Headlamp repository.

  4. Overview of Headlamp AI Assistant

    main

    The Headlamp AI Assistant is an alpha-state plugin that provides a conversational interface for Kubernetes cluster management. It is context-aware, using cluster information to help users manage resources, troubleshoot issues, and generate Kubernetes YAML manifests through natural language.

    Key Capabilities:

    • Conversational Management: Ask questions, get explanations, and issue commands.
    • Resource Generation: Generate manifests for deployments, services, etc.
    • In-depth Analysis: Diagnose issues and interpret logs.
    • Multi-Provider Support: Connect to various AI backends.
    • Configurable Tools: Enable or disable specific capabilities like direct Kubernetes API access.

  5. Overview of the Kubeflow Plugin for Headlamp

    main

    The Kubeflow plugin provides an operator-focused UI for managing and observing Kubeflow resources within Headlamp. It is designed to augment the Kubeflow Central Dashboard by providing deep, Kubernetes-centric observability, such as inspecting cluster health, pods, status conditions, and detailed configurations of Machine Learning workloads.

    Key Feature: Auto-detection The plugin uses a native probing design to dynamically detect available Kubeflow dependencies by checking the cluster APIs for Custom Resource Definitions (CRDs). This means you do not need a full Kubeflow installation; the plugin will only unlock the specific UI sections in the Headlamp sidebar that correspond to the components you have installed (e.g., only Katib or only Notebooks).

  7. Overview of the Kubeflow Spark Plugin

    main

    The Kubeflow Spark Plugin provides Kubernetes-native observability for Spark workloads within Headlamp. It is designed for platform operators, SREs, and ML engineers to monitor SparkApplication and ScheduledSparkApplication resources without using kubectl or external dashboards.

    Key Capabilities:

    • Cluster-wide Visibility: Unified dashboard for Spark applications across all namespaces.
    • Resource Health: Real-time status badges derived from operator conditions (e.g., Failed, Running, Pending).
    • Topology Inspection: View driver and executor pod relationships, node placement, and container status.
    • Schedule Management: Monitor recurring jobs with human-readable cron schedules and success rates.
    • Configuration Access: Inspect Spark configurations, environment variables, volumes, and RBAC roles.

  8. Overview of Cluster API Plugin features

    main

    The Cluster API Plugin provides a graphical interface for managing Kubernetes Cluster API (CAPI) resources within Headlamp. It is designed to simplify multi-cluster operations through visual relationship mapping and interactive controls.

    Key Capabilities

    • Map View Integration: Visualizes CAPI resource ownership relationships.
    • Direct Resource Scaling: Scale nodes directly via custom UI controls.
    • 1-Click Kubeconfig Download: Instantly fetch workload cluster configurations.
    • Rich Status Monitoring: Provides human-readable conditions (e.g., Ready, Paused, Error).
    • Resource Tracking: Maps machines to their parent resources and visualizes ClusterClass configurations.

    Compatibility

    • API Versions: Supports v1beta1 and v1beta2.
    • Providers: Provider-agnostic; works with Docker, AWS, Azure, GCP, vSphere, etc.

  9. Overview of the Kubeflow Pipelines Plugin

    main
    The Kubeflow Pipelines (KFP) plugin for Headlamp provides platform operators and SREs with Kubernetes-native visibility into the KFP control plane. Unlike the standard KFP UI, this plugin interacts directly with the Kubernetes API Server via KFP CRDs, meaning it does not require access to the KFP backend database (MySQL/PostgreSQL) or the KFP API service. It is designed for operational health monitoring, auditability, and root-cause discovery.

  10. Overview of the Kubeflow Katib Plugin

    main

    The Kubeflow Katib Plugin provides platform operators, SREs, and ML engineers with Kubernetes-native visibility into Kubeflow Katib resources directly within Headlamp. It surfaces hyperparameter tuning workflow data, including Experiment status, Trial metrics, and Suggestion assignments, reducing the need to switch between kubectl and other dashboards.

    Key operational insights provided:

    • Experiment Status: Track active, failed, or converged experiments.
    • Trial Monitoring: View running trials and their reported metric values.
    • Suggestion Tracking: Monitor requested vs. assigned suggestions.
    • RBAC Visibility: Inspect service accounts and worker type relationships.
    • Progress Tracking: Monitor experiment progress against trial budgets.

  11. Overview of the Headlamp Argo CD Plugin

    main

    The Headlamp Argo CD Plugin integrates Argo CD management directly into the Headlamp Kubernetes workflow. It allows users to view and manage Argo CD Application resources using standard Kubernetes RBAC instead of requiring a separate Argo CD session token or REST API access.

    Key Features:

    • Application List View: Displays all Argo CD Application resources with columns for project, source repo, target revision, sync status, and health status.
    • Application Detail View: Shows metadata including project, source, destination, sync status, health status, and Kubernetes events.
    • Sync Action: Triggers a synchronization by patching the Application's .operation field via the Kubernetes API.
    • Refresh Action: Requests a refresh (supporting normal and hard types) by setting the argocd.argoproj.io/refresh annotation.
    • Argo CD Sidebar Icon: Adds an official Argo CD icon to the Headlamp sidebar.

  12. Overview of kro Plugin features

    main

    The kro plugin provides specialized views for kro (Kube Resource Orchestrator) within Headlamp, utilizing Kubernetes watches for live updates. Key capabilities include:

    • ResourceGraphDefinition (RGD) Management: View lists and details of RGDs, including their state, conditions, topological resource order, and a flattened SimpleSchema summary.
    • Generated API Instance Views: Dynamically discovers CRDs generated by active RGDs. Provides detail pages for these instances containing conditions, spec, and standard Headlamp edit/delete actions.
    • Sub-resource Monitoring: Displays resources created by kro (identified via ownership labels) with health status, links to their native Headlamp pages, and environment-resolved values (e.g., PVC storageClassName, Deployment ready count, Service type).
    • Map Integration: A dedicated kro source for the Headlamp Map view that visualizes RGDs, instances, and kro-managed resources.
    • New Instance Creation: Provides a shortcut to Headlamp's YAML editor, pre-filled with a minimal valid instance based on the RGD's SimpleSchema.
    • Resilience: Includes graceful degradation, providing installation pointers if kro is missing from the cluster and per-section RBAC handling to prevent broken pages.