YLS Schema Loader

Finds the right yaml-language-server schema for Kubernetes-style YAML documents and helps you wire it up — without guessing repo layouts or digging through source by hand.

redhat.vscode-yaml (which bundles yaml-language-server) only validates a document once it's associated with a schema, either via a yaml.schemas setting or an inline # yaml-language-server: $schema=... comment. For a custom resource — an ArgoCD ApplicationSet, a Kargo Stage, or any other kind/apiVersion — finding that schema URL is manual, tedious work. This extension automates the lookup.

Features

• Detects kind/apiVersion in open YAML documents (including multi-document files) and checks whether a schema is already associated.
• Resolves a schema through four stages, first match wins:
  1. A bundled, curated mapping of common kinds (see below).
  2. An in-session cache of resolutions already made this window.
  3. A GitHub heuristic: guesses the source repo from the apiVersion group, probes common schema file paths, and falls back to a GitHub code search.
  4. A datreeio/CRDs-catalog lookup, tried only when step 3 finds no source-code repo (or none with a schema) and wasn't rate-limited — a single conventional URL, verified to exist before it's used.
• Never edits silently. An uncovered document gets an info diagnostic and a quick-fix — nothing is written until you act on it. A schema found via GitHub always shows you the candidate (or a picker, if there's more than one) and requires explicit confirmation before anything is inserted. Bundled-mapping and CRDs-catalog matches apply directly — both are already verified before being offered, so there's nothing ambiguous to confirm.
• Two ways to apply a fix: the quick-fix on the diagnostic, or the command palette — YLS Schema Loader: Insert Schema for Current Document — which resolves whichever document your cursor is in.

Bundled schemas

53 kind/apiVersion pairs across the following ecosystems, grouped by project:

• Kargo (kargo.akuity.io/v1alpha1) — all resource kinds: Stage, Warehouse, Promotion, ClusterConfig, ClusterPromotionTask, Freight, Project, ProjectConfig, PromotionRequest, PromotionTask, Target.
• ArgoCD ecosystem (argoproj.io/v1alpha1) — Argo CD (Application, AppProject, ApplicationSet), Argo Rollouts (Rollout, AnalysisTemplate, ClusterAnalysisTemplate, AnalysisRun, Experiment), Argo Workflows (Workflow, WorkflowTemplate, ClusterWorkflowTemplate, CronWorkflow), and Argo Events (EventBus, EventSource, Sensor).
• Flux — GitRepository, HelmRepository, OCIRepository, Bucket (source.toolkit.fluxcd.io/v1), Kustomization (kustomize.toolkit.fluxcd.io/v1), HelmRelease (helm.toolkit.fluxcd.io/v2), ImageRepository, ImagePolicy, ImageUpdateAutomation (image.toolkit.fluxcd.io/v1), Alert, Provider (notification.toolkit.fluxcd.io/v1beta3), Receiver (notification.toolkit.fluxcd.io/v1) — schemas sourced from the Flux-community-maintained flux2-schemas.
• Istio — VirtualService, Gateway, DestinationRule, ServiceEntry, Sidecar (networking.istio.io/v1), plus PeerAuthentication, AuthorizationPolicy, RequestAuthentication (security.istio.io/v1).
• Cilium (cilium.io/v2) — CiliumNetworkPolicy, CiliumClusterwideNetworkPolicy.
• cert-manager (cert-manager.io/v1) — Certificate, ClusterIssuer, Issuer.
• Kustomize (kustomize.config.k8s.io/v1beta1) — the native kubectl kustomize / kustomize CLI's own kustomization.yaml Kustomization kind, distinct from Flux's kustomize.toolkit.fluxcd.io Kustomization CRD above. Kustomize doesn't publish its own JSON schema, so this uses the community-maintained SchemaStore entry.
• External Secrets Operator (external-secrets.io/v1) — ExternalSecret.

Anything else falls through to GitHub discovery, or you can extend the mapping yourself with ylsSchemaLoader.additionalMappings.

Extension Settings

Requirements

• VS Code ^1.90.0.
• redhat.vscode-yaml is what actually validates against the schema once it's associated — this extension just does the lookup and wiring. It's not a hard dependency (the inline-comment insertionMode works without it), but you'll want it installed for the yamlSchemas insertion mode and for validation itself.

Known limitations

• No live-cluster CRD introspection — schemas come from the bundled mapping, your own overrides, or GitHub, not kubectl get crd.
• Only acts on documents with no existing schema association; it doesn't detect or offer to replace a stale one.
• GitHub discovery is a best-effort heuristic (a handful of conventional file paths, then a code search) — it won't find every schema in every repo layout. ylsSchemaLoader.githubRepoHints and additionalMappings are the escape hatches.

Development

npm install
npm run compile # type-check + compile src/ and test/ to out/
npm run bundle   # bundle src/extension.ts -> dist/extension.js (dev build)
npm test         # unit tests, then integration tests in a real VS Code instance
npm run lint     # eslint
npm run format   # prettier --write

The shipped extension is a single esbuild bundle (dist/extension.js, built automatically by vsce package/vsce publish via the vscode:prepublish script) — npm run compile's out/ output is only used for running tests, never for packaging.

Release Notes

See CHANGELOG.md.