Skip to content

ROADMAP.md

Latest commit

 

History

History
122 lines (96 loc) · 9.4 KB

ROADMAP.md

File metadata and controls

122 lines (96 loc) · 9.4 KB
Raw

KubeStellar Console Roadmap

This document outlines the planned direction for KubeStellar Console. It is a living document and will be updated as priorities evolve based on community feedback, user needs, and ecosystem changes.

Completed Milestones

v0.1 — Foundation (Q3 2025)

  • Multi-cluster dashboard with real-time health monitoring
  • Helm release tracking across clusters
  • Pod, deployment, and event monitoring cards
  • Demo mode with MSW mock data for offline usage
  • GitHub OAuth authentication
  • Dark/light theme support

v0.2 — Intelligence Layer (Q4 2025)

  • AI-powered missions system with Claude and kagent integration
  • Community missions browser with console-kb knowledge base
  • Contributor rewards system with leaderboard and coin economy
  • 80+ dashboard cards covering CNCF ecosystem
  • GPU monitoring cards (overview, inventory, utilization, reservations)
  • OPA, Kyverno, Falco, and Trivy security cards
  • ArgoCD application monitoring
  • Drag-and-drop dashboard customization with card catalog

v0.3 — Scale & Operations (Q1–Q2 2026)

  • Console Studio — Visual dashboard builder with AI card generation
  • Mission Control — Guided CNCF project deployment with Flight Plan blueprint, phased launch, and AI-assisted cluster assignment; dry-run mode and kind cluster E2E tests
  • Orbital Maintenance — Automated cluster maintenance missions with scheduling
  • Benchmark streaming — Real-time vLLM/llm-d performance data via Google Drive with hardware leaderboards
  • GPU namespace drill-down — Per-GPU-type, per-node allocation views
  • Workload import dialog — YAML, Helm, GitHub, and Kustomize import support
  • NPS survey system — In-app Net Promoter Score feedback collection
  • VCluster and KubeVirt cards for virtualized workloads
  • Marketplace — Community card preset marketplace with 45+ CNCF project templates
  • OpenSSF Scorecard improvements — Signed releases, SLSA provenance, scoped workflow permissions
  • 160+ total dashboard cards
  • Nightly and weekly automated releases with Helm OCI chart publishing
  • Comprehensive Auto-QA workflows for code quality, governance, and UI consistency
  • Contributor leaderboard with GitHub-synced rewards
  • AI Missions UX — Message edit/resend, microphone input, scroll-to-bottom, draft click-to-open, history toggle panel, mission sort by activity, retry on failure, response cancellation
  • Auth hardening — GA4 telemetry on auth failure paths (SSE 401, WS token missing, agent token failure, session refresh), agentFetch migration for all kc-agent calls, HS256-only JWT parsing (TAG-Security fix)
  • kc-agent API expansion — /nvidia-operators, /events/stream SSE, /federation/detect, agent token bridging to frontend
  • Responsive container-query rollout — Phase 3a/3b across 63 files: responsive skeleton grids, flex-wrap in CNCF status cards
  • Test infrastructure — Coverage from 0% to 91%: 10,000+ unit tests, 12-shard parallel coverage, coverage regression guard with auto-issue, post-merge Playwright verification against production
  • Code quality automation — UI/UX standards scanner with Storybook and Playwright visual regression, post-build vendor safety checks, MSW catch-all for unmocked routes
  • Backend refactoring — Monolith splits: sqlite.go (3,321 → 8 files), server_http.go/server_ai.go/server_operations.go into domain handlers, CardWrapper.tsx into 4 sub-components; 609 fmt.Sprintf calls converted to structured slog fields
  • ArgoCD ApplicationSet integration with security fixes
  • Saved Filter Sets — Snapshot all filters into named presets; merged Project Selector and Filter Panel into single dropdown
  • Learn dropdown — Auto-populated from YouTube playlist with video tutorials
  • Claude Code GitHub Action — AI-assisted PR review and issue triage via Claude Opus 4.6

v0.4 — AI-Native Observability (Target: Q3 2026)

This milestone crystallizes the near-term roadmap items into a cohesive theme: establishing KubeStellar Console as the canonical AI/ML workload visibility and operations layer for Kubernetes.

Core Scope

  • llm-d stack monitoring — First-class support for llm-d inference serving: EPP routing, model endpoint health, autoscaler status, disaggregated serving topology
  • Drasi reactive pipelines — Real-time change-feed dashboard for Drasi continuous queries, sources, and reactions across deployment modes (drasi-server, drasi-platform, CRD-based)
  • kagent/kagenti integration — Full agent lifecycle management through MCP-compatible interfaces

Quality & Testing

  • Nightly E2E expansion — Automated end-to-end testing across all 8 llm-d deployment guides on OpenShift
  • Marketplace v2 — Require live data hooks, unified controls, demo data, and install links for all card presets; community review process

UX & Accessibility

  • i18n completeness — Eliminate all hardcoded English strings; prepare for community localization contributions
  • Accessibility audit — Replace remaining window.confirm() dialogs, add ARIA labels, keyboard navigation for all interactive elements
  • GA4 UX funnel — Measure conversion from landing to agent install to first mission; identify and fix drop-off points
  • Component consistency — Migrate remaining raw HTML elements to shared UI components (Button, Modal, Dialog); standardize modal visibility patterns

Community Health

  • Adopters program — Populate ADOPTERS.MD with confirmed production users; define maturity tiers (install-mission vs. production deployment)
  • Contributor onboarding — Establish PR triage SLA, define ai-needs-human escalation path, and publish contributor guide update
  • Adoption metrics — Replace all TBD fields in docs/adoption-metrics.md with real measurements before any CNCF application

Near-Term (Q2–Q3 2026)

Mid-Term (Q3–Q4 2026)

  • GitOps integration milestone — First-class Flux + Argo CD support with observability parity, declarative Console configuration, and Mission Control deep links; see docs/plans/GITOPS-INTEGRATION-RFC.md
  • Multi-tenant RBAC — Role-based access control for teams sharing a Console instance, with namespace-scoped permissions
  • Plugin architecture — Extensible card and mission system allowing third-party developers to build custom dashboard components
  • Helm operator — Kubernetes operator for fleet-wide Console deployment and lifecycle management
  • Enhanced AI missions — AI-assisted troubleshooting missions that diagnose cluster issues and suggest remediation steps
  • Offline/air-gapped mode — Full Console functionality without internet connectivity for restricted environments
  • CNCF incubation preparation — Governance documentation, adopters program, and community growth metrics; target Q4 2026 TOC application
  • Third-party security audit — Engage CNCF-sponsored auditors (e.g., ADA Logics) for a formal code security audit; required gate for CNCF incubation
  • Multi-model AI backend — Support for multiple LLM providers (OpenAI, Ollama, vLLM) behind a unified mission interface, reducing vendor lock-in
  • Webhook-driven card updates — Push-based card refresh via Kubernetes webhooks instead of polling, reducing API server load on large clusters
  • Custom alert rules — User-defined threshold alerts on any card metric, with notification channels (Slack, email, PagerDuty)

Long-Term (2027+)

  • Policy engine — Built-in policy authoring, testing, and enforcement with OPA/Gatekeeper integration
  • AI-assisted operations — Proactive anomaly detection, capacity planning, and automated incident response via MCP
  • Federation — Console-to-Console federation for organizations managing multiple Console instances across regions
  • Compliance dashboards — Automated compliance reporting against CIS benchmarks, SOC 2, and HIPAA requirements
  • Collaborative dashboards — Real-time multi-user dashboard editing with presence indicators and conflict resolution
  • Workflow automation — Visual workflow builder for multi-step cluster operations (rolling upgrades, canary deployments, disaster recovery runbooks)
  • Embedded terminal — In-browser kubectl/helm terminal with context-aware autocomplete, scoped to the user's RBAC permissions

Non-Goals

KubeStellar Console intentionally does not aim to:

  • Replace kubectl — Console is a visual companion, not a CLI replacement. Power users should continue using kubectl, helm, and other CLI tools directly.
  • Be a general-purpose IDE — While Console includes AI-powered features, it is not a code editor or development environment.
  • Manage non-Kubernetes workloads — Console focuses exclusively on Kubernetes clusters and cloud-native workloads.
  • Provide its own container runtime — Console observes and manages existing clusters; it does not provision infrastructure.
  • Compete with commercial APM tools — Console provides operational visibility, not deep application performance monitoring. Use Datadog, New Relic, or Grafana for APM.

How to Influence the Roadmap

We welcome community input on priorities: