---

name: backend-and-configuration description: "Operate CubeStudio backend lifecycle, configuration overlays, RBAC, services, and frontend build customization." metadata: disco-role: operating disable-model-invocation: true license: NOASSERTION

CubeStudio backend and configuration

Use this sub-skill for CubeStudio backend maintenance tasks that involve Flask AppBuilder startup, config/project overlays, authentication and RBAC, backend view/API registration, Celery and watcher services, or frontend build/proxy customization.

Read first

1. references/config-reference.md for the mandatory runtime overlay model: the checked-in myapp/config.py and myapp/project.py placeholders are empty, while Docker and Kubernetes mount real overlay files at runtime.
2. references/backend-lifecycle.md for app startup, DB migration/init order, AppBuilder permissions, auth hooks, and Celery/watch service behavior.
3. references/frontend-build.md for the three frontend packages, safe proxy edits, and build commands that require explicit permission.
4. references/troubleshooting.md when imports, login, permissions, DB/Redis, Celery, watchers, or frontend proxy/builds fail.

For a safe static inspection of a provided CubeStudio-like checkout, run:

python scripts/inspect_cube_studio_structure.py /path/to/cube-studio --json

The helper reads files only. It does not import myapp, connect to DB/Redis, start services, run Docker, run Kubernetes commands, install npm packages, or build frontend bundles.

Use this sub-skill when the task asks to

• boot, inspect, or change the Flask AppBuilder backend lifecycle;
• decide where to edit config.py, project.py, or runtime environment variables for Docker/Kubernetes/local inspection;
• add or debug backend model-view/API registrations and FAB permissions;
• customize login/auth, user roles, header/JWT behavior, or project messaging hooks;
• reason about Celery worker/beat tasks or watch_workflow/watch_service;
• change setupProxy.js, frontend package scripts, or SPA build/deploy behavior.

Route elsewhere

• Raw Docker Compose or Kubernetes deployment, image registry rewrites, CRD ordering, namespace/secret/PVC setup: use deploy-and-operate.
• Notebook/resource group/GPU selectors, image catalog and online image builds: use compute-notebooks-and-images.
• Pipeline DAGs, task templates, Argo workflow generation, run history: use pipelines-and-job-templates.
• Dataset, metadata, dimension, SQLLab, ETL details: use data-metadata-and-sqllab.
• Model registry, inference services, AIHub, chat, LLM gateway behavior: use serving-aihub-and-llm.

Keep domain-specific model/view fields in those sibling sub-skills; use this sub-skill only for the shared backend plumbing that makes those domains available.