Skip to content
GreenKube

Web UI

The GreenKube UI is a single-page application served by the API at /. Every page reads the API with your session; no page starts a scan or writes anything on load. Status is always written as a word (Measured, Projected, Action needed, Failing), never shown by colour alone. Try it without installing anything at demo.greenkube.cloud.

GreenKube Overview: measured allocated savings, savings pending verification, the projected opportunity, changes in flight and what to do next (demo data)

PagePurpose
Overview (/)Savings first: measured allocated savings with the realization factor, savings pending verification, the projected opportunity, changes in flight, and the top recommendations
Recommendations (/recommendations)The ranked list with filters and plan preview. Each detail page shows the proposal, expected value, delivery eligibility with blockers and fixes, projected impact by component, released capacity, consensus across sources, evidence, pull requests, measurement and lifecycle timeline
Changes (/changes)One board from approved plan to measured outcome: Plan → PR open → Merged → Synced → Applied → Measuring → Outcome. Rows that need a human are flagged (sync failed, pull request failed, regressed)
Ledger (/ledger)Savings totals per method and every applied change with its measurement, confidence interval, original projection and split across recommendations
Sources (/sources)Every recommendation source: status, last run, counts, measured track record and recent runs
Setup (/setup)The setup checklist: the four capabilities (discovery, delivery, sync tracking, measurement), each check with its fix, and the repositories found by discovery
Footprint (/footprint)Cost, energy and CO₂e per workload, namespace and node, with the allocated totals, the node idle share and the cluster total side by side, and Nodes and Report sections (CSV and JSON exports)
Automation (/automation)GitOps discovery, workload binding coverage and manual resolution
Settings (/settings)Account, sessions, API keys and users

The ranked list shows, on each card, the projected CO₂e and cost savings, the capacity it frees, its evidence, risk, confidence and expected value, with filters by namespace, source, risk and capability, Rescan, Open PR and Apply manually. Rightsizing (CPU and memory, decreases and reliability increases) and orphaned LoadBalancer and PersistentVolume deletions are the only types.

GreenKube Changes board: every change from plan to pull request, merge, sync, rollout and measured outcome (demo data)

The change page shows the applied change, its measurement and the released capacity:

GreenKube Change page: a measured change with its allocated saving, 90% interval and released capacity (demo data)

See Footprint for the idle share, the node badges and the report exports.

GreenKube Footprint, Nodes section: node inventory with CPU and memory capacity (demo data)

GreenKube Footprint, Report section: configure, preview and export the cost, energy and CO₂e report as CSV or JSON (demo data)

Projected, allocated and realized are never summed

Section titled “Projected, allocated and realized are never summed”

Every page uses one vocabulary and shows each kind of saving in its own panel, with its method as a label:

  • Measured allocated savings: what the changed workloads stop being charged, measured after the change and accrued in the ledger.
  • Realization factor: the expected share of an allocated saving that reaches the bill, labelled as an expectation.
  • Realized (billed) savings: shown only when savings measured on the bill exist. The word “Realized” is reserved for them; a measurement outcome reads Saving measured or No saving measured.
  • Pending verification (Projected): the yearly projection of applied changes not measured yet.
  • Opportunity (Projected): the projected yearly savings of open recommendations. Reliability recommendations add resources on purpose and are counted apart, never as savings.
  • User declared (Ledger only): savings declared on manual applies, never counted as measured.

With accounts (session, proxy or oidc mode) you sign in with a username and password or through single sign-on. A headless install in api_key mode asks for an API key instead and keeps you signed in with a signed cookie (HTTPS or localhost only); see Installation: SSO and accounts.

Pages show only the actions the signed-in user may perform. Opening a pull request, ignoring, restoring and resolving a binding need the write scope; approving a plan needs approve. A short Read-only access note says why actions are missing, and the server enforces the same scopes.

With config.externalUrl set, pull requests link back to the recommendation (/recommendations/{id}), the change (/changes/{plan_id}) and its ledger entry (/ledger?change=…).