Kubernetes Helm Chart

Influxion has one supported Helm chart: helm-chart.

The chart deploys the FastAPI app and, when enabled, Celery workers as sidecar containers in the same pod. This is intentional: generated PDFs are stored in a short-lived pod emptyDir so the worker that renders the file and the app that serves the download must share the same runtime directory.

Main resources

  • Deployment: one app container plus worker.count Celery sidecars.

  • Service: exposes the app container on the configured HTTP port.

  • Secret: stores database URL, RabbitMQ broker URL, Entra ID, and Key Vault values.

  • StatefulSet + Service: optional built-in Postgres.

  • Deployment + Service: optional built-in RabbitMQ.

  • Ingress: optional public route with cert-manager issuer annotation.

  • ClusterIssuer: optional chart-managed Let’s Encrypt issuer.

Important values

image:
  app:
    repository: ghcr.io/nedomkull-mathematical-modeling/influxion-app
    tag: "2026.11.19"
  worker:
    repository: ghcr.io/nedomkull-mathematical-modeling/influxion-worker
    tag: "2026.11.19"

imagePullSecrets: []

app:
  replicaCount: 1

worker:
  enabled: true
  count: 3

auth:
  tenantId: "<tenant-guid>"
  apiClientId: "<api-client-guid>"

vault:
  url: "https://<vault>.vault.azure.net"
  clientSecret: "<client-secret>"

ingress:
  enabled: true
  className: nginx
  hosts:
    - host: influxion.example.com
  tls:
    enabled: true
    secretName: influxion-tls

certManager:
  enabled: true
  clusterIssuerName: letsencrypt-prod

app.replicaCount must stay at 1 while the runtime PDF store is pod-local. Scaling the app horizontally would allow a download request to land on a pod that does not have the generated file.

Container registries

The chart defaults to GitHub Container Registry image repositories: ghcr.io/nedomkull-mathematical-modeling/influxion-app and ghcr.io/nedomkull-mathematical-modeling/influxion-worker. Override image.app.repository and image.worker.repository for Azure Container Registry or local minikube images.

For private GHCR packages, create a Kubernetes Docker registry secret and set imagePullSecrets:

kubectl create secret docker-registry ghcr-pull-secret \
  --namespace influxion \
  --docker-server=ghcr.io \
  --docker-username=<github-username> \
  --docker-password=<github-token>
imagePullSecrets:
  - name: ghcr-pull-secret

Local Kubernetes

To start Influxion locally in Kubernetes, use minikube and the local deployment script:

minikube start --driver=docker
minikube addons enable ingress

.\deploy_to_docker.ps1 `
  -AzureApiClientId "<api-client-id>" `
  -AzureTenantId "<tenant-id>" `
  -IngressHost "influxion.local"

The script builds influxion-app:2026.11.19 and influxion-worker:2026.11.19 inside minikube’s Docker daemon and deploys the chart using those local image names.

After deployment, either add <minikube-ip> influxion.local to your hosts file or port-forward the service:

kubectl port-forward -n influxion svc/influxion-influxion-app 8000:80

Let’s Encrypt

The supported AKS script installs cert-manager and ingress-nginx, waits for their deployments, creates a letsencrypt-prod ClusterIssuer using HTTP-01, and then deploys the chart with:

--set ingress.enabled=true `
--set-string ingress.className="$IngressClass" `
--set-string ingress.hosts[0].host="$IngressHost" `
--set ingress.tls.enabled=true `
--set-string ingress.tls.secretName="influxion-tls" `
--set-string certManager.clusterIssuerName="letsencrypt-prod"

The chart adds the matching cert-manager.io/cluster-issuer annotation to the Ingress. Once DNS points at the ingress-nginx external IP, cert-manager can solve the HTTP-01 challenge and populate the TLS secret.

Validation

Run these checks before deploying changes:

helm lint helm-chart
helm template influxion helm-chart --debug

For AKS-style rendering, include the same image, auth, vault, ingress, and cert-manager overrides used by deploy_to_aks.ps1.