> ## Documentation Index
> Fetch the complete documentation index at: https://engineering.clinikally.work/llms.txt
> Use this file to discover all available pages before exploring further.

# Deployment and operations

> How Clinikally services are packaged, deployed, and observed.

## Deployment model

Runtime services are containerized and deployed with Helm. Service configuration is organized by environment:

```text theme={null}
DevOps/helm-charts/services/values/
├── DEVELOPMENT/
├── STAGING/
└── PRODUCTION/
```

Shared Helm templates are in `devops/helm-charts/services/templates/`:

* `deployment.yaml`
* `service.yaml`
* `cronjob.yaml`
* `keda-scaledobject.yaml`

The values tree contains service-specific releases such as Atlas, Compass, Dermadesk, Estimo, Gearbox processes, Hodor, Konsilo, Medusa components, Mint, Notify, Orderbox API/worker/cronjobs, Reflekt, Sesh, Storehouse API/worker, Tetris, and Vector API/worker.

## Operational dependencies

| Concern                        | Platform evidence                                    |
| ------------------------------ | ---------------------------------------------------- |
| Secrets                        | AWS Secrets and CSI driver configuration             |
| Autoscaling                    | KEDA ScaledObject templates/configuration            |
| Metrics and tracing            | Datadog, Prometheus, and Grafana chart configuration |
| Logs                           | Loki/Alloy configuration                             |
| Database connection pooling    | PgBouncer chart/configuration                        |
| Progressive/traffic operations | Argo CD and Terraform/traffic-split scripts          |

## Release checklist

1. Confirm the service’s active code branch and the intended environment values file.
2. Review required secrets and new configuration keys; never introduce plaintext credentials in Helm values.
3. Verify database migration ownership and backwards compatibility.
4. For async changes, provision queues/DLQs, retry policy, dashboards, and alerting before enabling producers.
5. Validate readiness/liveness and autoscaling settings for API and worker processes independently.
6. Confirm observability tags and correlation IDs are emitted for cross-service requests/events.
7. Use a staged rollout and preserve an explicit rollback target.

## Incident triage order

For a customer transaction issue, inspect in this order: Hodor/session context → Vector/Dermadesk request → Medusa cart/order → Mint gateway/callback → Orderbox worker/events → Storehouse reservation → Notify delivery. Use the correlation identifiers propagated by the participating services rather than timestamps alone.