nvcf-explore-stack

par nvidia

Naviguer et expliquer la pile auto-hébergée NVCF dans le monorepo. Mappe les releases helmfile à leurs charts, sous-arbres de sources d'images, hooks helm, namespaces, et…

npx skills add https://github.com/nvidia/nvcf --skill nvcf-explore-stack

NVCF Explore Stack

Help a user or developer navigate the NVCF self-hosted stack inside the monorepo. Identify what a release is, what installs it, what it depends on, and which subtree owns the chart and the runtime image.

Instructions

This skill is for both NVCF users and NVCF developers. Users use it to understand stack dependencies, deployment order, and CI/CD migration questions. Developers use it to find the owning chart, source subtree, hook, or helmfile stage before making changes.

Use this skill long enough to answer the question, then hand off to the right execution skill. Always cite the source file path so the user can verify.

Required inputs

Read these from the monorepo root.

Authoritative (always read first when answering):

  • deploy/stacks/self-managed/helmfile.d/00-observability-infrastructure.yaml.gotmpl
  • deploy/stacks/self-managed/helmfile.d/01-dependencies.yaml.gotmpl
  • deploy/stacks/self-managed/helmfile.d/02-core.yaml.gotmpl
  • deploy/stacks/self-managed/helmfile.d/03-observability.yaml.gotmpl
  • deploy/stacks/observability/helmfile.d/01-observability.yaml.gotmpl
  • deploy/stacks/nvcf-compute-plane/helmfile.d/01-dependencies.yaml.gotmpl
  • deploy/stacks/nvcf-compute-plane/helmfile.d/02-nvca.yaml.gotmpl

Chart-level (when the chart is checked into the monorepo):

  • deploy/helm/<chart>/Chart.yaml
  • deploy/helm/<chart>/values.yaml

Common questions

What deploys X : Look up release X in the helmfile stage files. Return chart name, version, namespace, and which gotmpl file declares it. If the chart is checked in, also point at deploy/helm/<chart>/.

What does X depend on : Return the needs: chain for that release plus the stage gate it sits behind (control-plane stages 0 -> 1 -> 2 -> 3, then compute-plane). Include any profile, condition:, or component mode that gates whether X deploys at all.

What hooks run for X : Read the checked-in chart under deploy/helm/<chart>/ when available. Search its templates/ directory for Helm hook annotations, weights, hook events (pre-install / post-install), images used, and purpose. Cite the chart-relative template file path. If the chart is not checked in, cite its Helmfile chart reference and state that local templates are unavailable.

Walk me through the full deployment order : Summarize control-plane stages 0 through 3 from the self-managed gotmpl files. Stage 0 delegates to the shared observability Helmfile when observability.profile is enabled. Stage 3 installs State Metrics and the function autoscaler for control and all. Then summarize the compute-plane stage from deploy/stacks/nvcf-compute-plane/helmfile.d/01-dependencies.yaml.gotmpl and deploy/stacks/nvcf-compute-plane/helmfile.d/02-nvca.yaml.gotmpl. Call out which releases run in parallel inside a stage and which are serialized by needs:.

Which subtree do I edit to change X : Point at the Helmfile path for orchestration, deploy/helm/<chart>/ for checked-in chart wiring, and the image source under src/, infra/, or migrations/ for runtime behavior. If a referenced chart is not checked in, report its Helmfile chart reference and do not claim a local source path.

What namespaces does the stack use : Return the list from the helmfile (namespace: per release).

Subtree mapping

The stack lives in three layers across the monorepo:

ConcernLives at
Helmfile orchestration (stage ordering, env wiring, secrets flow)deploy/stacks/self-managed/
Chart manifests, helm hooks, valuesdeploy/helm/<chart>/ when checked in; otherwise use the Helmfile chart reference
Runtime application code, migrationssrc/, infra/, migrations/

When a question crosses layers, answer by layer and tell the user the order to edit (chart wiring first if the deploy contract changes, image source if behavior changes).

Tone

Assume the user is onboarding to the stack. Be concise. Always include the chart name, version, and the gotmpl path when referencing a release. Prefer one short paragraph plus a code-block citation over prose.

Skill handoff candidates

After exploring, suggest the next skill when applicable:

  • nvcf-self-managed-installation for installing, upgrading, or tearing down the stack
  • docs/dev/local-development.md for k3d / local cluster work
  • nvcf-self-managed-cli for nvcf-cli usage against an installed stack
  • docs/AGENTS.md and fern/versions/main.yml for routing the user to a published docs page
  • tools/ci/check-doc-version-sync for keeping the documentation manifest in sync with the docs version catalog

Plus de skills de nvidia

compileiq-debug
nvidia
Utilisez quand quelque chose ne va pas : Search() bloque, toutes les évaluations retournent INVALID_SCORE, les scores ne s'améliorent pas, chaque configuration retourne le même nombre, erreurs ptxas…
create-github-pr
nvidia
Créer des pull requests GitHub en utilisant l'interface en ligne de commande gh. Utiliser lorsque l'utilisateur souhaite créer une nouvelle PR, soumettre du code pour révision, ou ouvrir une pull request. Mots-clés de déclenchement -…
nemoclaw-maintainer-cross-issue-sweep
nvidia
Analyse les autres problèmes ouverts pour trouver ceux qu’une PR donnée pourrait également corriger ou casser accidentellement. Génère des opportunités de correctifs adjacents et des risques de contradiction avec fichier:ligne…
fhir-basics
nvidia
Apprend aux agents comment fonctionnent les API FHIR R4, quelles ressources sont disponibles, comment les interroger avec des paramètres de recherche, et comment analyser correctement tous les formats de réponse…
compileiq-validate-result
nvidia
Utiliser APRÈS qu'une recherche soit terminée et AVANT de réclamer un accélérateur ou d'expédier un ACF. Charge le CSV dump_results, extrait les K meilleurs candidats (mono-objectif)…
changelog-audit
nvidia
Auditer le CHANGELOG.md de Warp avant une publication : récupérer les entrées perdues, trier par impact utilisateur, affiner le langage des entrées, ajuster les retours à la ligne et (en mode branche de publication) mettre à jour la comparaison…
maintain-dynamic-plugins
nvidia
Maintenir les chargeurs de plugins dynamiques NeMo Relay, les manifestes, les SDK natifs Rust, le protocole worker gRPC, le SDK worker Python, la documentation, les tests et la couverture du workflow de publication
dgx-diagnose
nvidia
Diagnostiquer les problèmes courants du DGX Station GB300 — plantages CUDA, ciblage incorrect du GPU, bugs de conteneur vLLM/SGLang, problèmes d'état MIG, erreurs NVLink/Fabric Manager,…