| name | v9-deployment-verifier |
| description | Verify V9 deployment readiness and deployment regressions after changes to install scripts, Ansible roles, Helm charts, smoke checks, or middleware topology. Use when Codex modifies files under `install.sh`, `ansible/`, `helm-chart/`, `images/load-images.sh`, `foundation/`, `init-scripts/`, or deployment validation scripts and needs to run the standard local validation flow. |
V9 Deployment Verifier
Run the repository validator after every deployment-related change:
python scripts/validate_deploy.py --mode regression
Use release mode before claiming the package is customer-installable:
python scripts/validate_deploy.py --mode release
What the validator covers
- Required deployment files exist.
install.sh still points to existing companion scripts.- Helm lint and template succeed for
v9-infra, core-agent-service, v9-services, and the umbrella chart.
v9-infra renders all four middleware components.core-agent-service still renders POSTGRES_URL, RABBITMQ_URL, REDIS_URL, and OPENSEARCH_URL.- Ansible syntax is checked when
ansible-playbook is installed.
- Missing local tools such as
ansible-playbook, kubectl, docker, or bash are reported explicitly as skipped runtime checks.
How to interpret results
- Treat any
FAIL as a real regression or delivery blocker.
- Treat
WARN in regression mode as a follow-up item, not as proof of deployability.
- Treat
SKIP as "not verified on this machine"; mention the missing tool in the final response.
- Do not say the deployment is fully verified if release mode fails or if runtime checks are skipped.
Guardrails
- Prefer the validator script over ad hoc one-off checks.
- If you change deployment logic, run the validator before replying.
- If the validator cannot cover a runtime path because the machine lacks tooling or cluster access, say exactly which checks were skipped.
- Keep the validator script updated when deployment entrypoints or expected middleware components change.
