refactor(deploy)!: rename mesh to studio in helm, compose and images

[viktormarinho]

Summary

De-mesh task 09 (deploy / infra). The chart name (chart-deco-studio) and image names (ghcr.io/decocms/studio/*) were already renamed in the repo-rename pass, so this PR finishes the remaining "mesh" identifiers in deploy/, tests/multi-pod and the CI wiring that references them.

Out of scope (owned by other open de-mesh PRs / task 10): apps/mesh directory paths, MESH_* env vars in tests/resilience (#4645), @decocms/mesh npm package + release-mesh.yaml / publish-mesh-sdk-npm.yaml workflow names (#4653 / task 10).

External-contract renames — ops coordination required

Contract	Old	New	Coordination requirement
chart-deco-studio values key	configMap.meshConfig	configMap.studioConfig	Breaking. Rename the key in every environment values file before upgrading to chart 0.13.0. A template fail-guard rejects the old key with an explicit message, so an un-migrated upgrade fails loudly at render time instead of shipping an empty ConfigMap.
chart-deco-studio dispatch-role env	MESH_DISPATCH_ROLE	STUDIO_DISPATCH_ROLE (canonical) + MESH_DISPATCH_ROLE kept as alias	None yet — deliberately dual-set. The chart sets both names to the same value, so it works with images both before and after #4645 (the STUDIO_* fallback). No image/chart ordering constraint. Follow-up: drop the MESH_ alias once all deployed images read STUDIO_*. Both names are now reserved (chart rejects them in env / worker.env).
sandbox-env values block	mesh.* (mesh.namespace, mesh.serviceAccountName, mesh.serviceName, mesh.servicePort, mesh.podSelectorLabels)	studio.*	Breaking. Update values files / --set mesh.* flags to studio.* before upgrading to chart 0.10.0. Fail-guard rejects the old block. The release-sandbox-charts.yaml install snippet now prints --set studio.*.
sqlite volume path	/app/data/mesh.db	unchanged	Intentionally kept: it is a persisted file on existing volumes; renaming it would orphan data on running single-node deployments. A template comment documents this.
Chart names / release names / image names	chart-deco-studio, ghcr.io/decocms/studio/{studio,studio-nginx,s3-sync}	unchanged	Already renamed in a previous pass — no action.
multi-pod compose services	mesh-1/2/3	studio-1/2/3	Test-infra only, no running-cluster impact.

Chart version bumps: chart-deco-studio 0.12.2 → 0.13.0, sandbox-env 0.9.5 → 0.10.0 (both breaking-values releases).

Changes

• deploy/helm/studio: meshConfig → studioConfig across values/templates/README; dual STUDIO_DISPATCH_ROLE/MESH_DISPATCH_ROLE env in API + worker deployments; reserved-var validation covers both names; fail-guard for the old values key.
• deploy/helm/sandbox-env: mesh.* → studio.* values block (+ fail-guard); prose/comments de-meshed in templates, values, README, examples, housekeeper script; Chart description updated.
• deploy/helm/sandbox-operator: README prose.
• tests/multi-pod: compose services/anchors/POD_NAME mesh-N → studio-N; PODS.MESH_N → PODS.STUDIO_N, PodName type and scenario literals renamed (lexicographic leader ordering studio-1 < studio-2 < studio-3 is preserved, so projector-parity assumptions hold).
• .github/workflows: multi-pod.yml log-dump service names; release-sandbox-charts.yaml install snippet.
• tests/resilience needed no changes (its compose has no mesh service names; MESH_CLUSTER_URL there belongs to refactor: rename MESH_* env vars to STUDIO_* with fallback #4645).

Testing

• helm lint + helm template on both charts (default, postgres, and examples/values-kind.yaml values) — render OK; both fail-guards and the reserved-var check verified to fail with the intended messages; rendered manifests show both dispatch-role vars.
• docker compose -f tests/multi-pod/docker-compose.yml config -q and ... tests/resilience/docker-compose.yml config -q — OK.
• bun run fmt (clean), bun run lint (0 errors), bun run check (all workspaces pass).
• Multi-pod/resilience docker suites not run locally (Docker-compose build of the full app); the renames there are mechanical and covered by CI's multi-pod job.

🤖 Generated with Claude Code

---

Summary by cubic

Renamed remaining "mesh" identifiers to "studio" across Helm charts, Docker Compose (multi-pod), and CI. Bumps chart-deco-studio to 0.13.0 and sandbox-env to 0.10.0 with fail-guards; update values before upgrading.

• Migration

  • chart-deco-studio: rename configMap.meshConfig → configMap.studioConfig (templates hard-fail on the old key). Chart sets both STUDIO_DISPATCH_ROLE (canonical) and MESH_DISPATCH_ROLE (alias); both names are reserved in env and worker.env. SQLite path stays /app/data/mesh.db.
  • sandbox-env: rename mesh.* → studio.* (namespace, serviceAccountName, serviceName, servicePort) before upgrading; the chart fails if mesh.* is present. Release snippet now prints --set studio.*.
  • Scripts/ops: multi-pod services are now studio-1/2/3.

• Refactors

  • deploy/helm/studio: values/README/templates moved to studioConfig; explicit fail on configMap.meshConfig; API and worker deployments set STUDIO_DISPATCH_ROLE + MESH_DISPATCH_ROLE; reserved-var checks updated.
  • deploy/helm/sandbox-env: prose/comments de-meshed across templates, values, README, examples, housekeeper; RBAC adds a fail-guard if .Values.mesh is set; preview Gateway docs clarified.
  • CI/tests: mesh-1/2/3 → studio-1/2/3 in compose, constants, scenarios; workflow log-dump targets updated.

^{Written for commit c35dc4d. Summary will update on new commits.}