Installation / Deployment

This guide summarizes the supported ways to run Influxion and the required configuration.

Prerequisites

  • Register the Influxion API in Entra ID before running any deploy scripts. You must know the API App Registration “Application (client) ID” and provide it as INFLUXION_AZURE_API_CLIENT_ID.

  • See docs/azure-setup.rst for details on App Registration, scopes, and Key Vault.

  • Common environment variables (set as script arguments or via environment): - INFLUXION_AZURE_TENANT_ID (required) - INFLUXION_AZURE_API_CLIENT_ID (required) - INFLUXION_VAULT_URL (optional) - INFLUXION_AZURE_API_CLIENT_SECRET (optional; used by the service to talk to Azure, if not using managed identity)

For client scripts that call the API, set:

  • INFLUXION_CLIENT_ID

  • INFLUXION_CLIENT_SECRET

Direct on Ubuntu (Supervisor + NGINX)

Standalone (API only): deploy_ubuntu_standalone.sh

  • Minimal footprint, can run on Azure B1s (1 vCPU/1 GiB). Similar sizes work on AWS (e.g., t3.micro) and GCP (e.g., e2-micro).

  • Example:

    ./deploy_ubuntu_standalone.sh \
      --postgres-password "<pg-pass>" \
      --azure-tenant-id "<tenant-guid>" \
      --azure-api-client-id "<api-client-guid>" \
      --ingress-host "influxion.example.com" \
      --acme-email "you@example.com"
    

Distributed (API + Celery workers): deploy_ubuntu.sh

  • Suitable for a small VM with more headroom than standalone. Provides possibility of asynchnonous PDF generation with Celery workers.

  • Example:

    ./deploy_ubuntu.sh \
      --postgres-password "<pg-pass>" \
      --rabbitmq-password "<rmq-pass>" \
      --azure-tenant-id "<tenant-guid>" \
      --azure-api-client-id "<api-client-guid>" \
      --ingress-host "influxion.example.com" \
      --acme-email "you@example.com"
    

Local with Docker Compose

A .env file at the repo root can provide:

  • INFLUXION_AZURE_TENANT_ID

  • INFLUXION_VAULT_URL

  • INFLUXION_AZURE_API_CLIENT_ID

  • INFLUXION_AZURE_API_CLIENT_SECRET

Standalone: docker-compose.standalone.yml

  • Starts Postgres and the API only.

  • Command:

    docker-compose -f docker-compose.standalone.yml up --build
    

Distributed: docker-compose.yml

  • Starts Postgres, RabbitMQ, API, and multiple workers.

  • Command:

    docker-compose up --build
    

Both Compose setups include a Postgres service and set INFLUXION_DATABASE_URL for the app/workers.

Local Kubernetes (minikube)

Use deploy_to_docker.ps1 to provision to local Kubernetes (minikube) with Helm. The script starts minikube, enables the ingress addon, builds the app and worker images inside minikube’s Docker daemon, and deploys helm-chart with local image repositories.

Start minikube directly if you want to run the steps yourself:

minikube start --driver=docker
minikube addons enable ingress

Then deploy Influxion:

  • Example (PowerShell):

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

If you use the ingress hostname, point it at the minikube IP:

minikube ip

Add the resulting IP to your hosts file:

<minikube-ip> influxion.local

You can also skip local DNS and port-forward the service:

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

Then open http://localhost:8000/docs or http://localhost:8000/swagger.

Azure Kubernetes Service (AKS)

Use deploy_to_aks.ps1 to deploy the supported helm-chart. The script installs cert-manager and ingress-nginx, creates the Let’s Encrypt ClusterIssuer, and deploys Influxion with an HTTPS Ingress.

Azure Container Registry

ACR remains the default registry mode. The script creates the registry if needed, logs in, builds and pushes app/worker images, attaches the ACR to AKS, and points Helm at the ACR image repositories.

  • Example (PowerShell):

    .\deploy_to_aks.ps1 `
      -ResourceGroupName "my-rg" `
      -AKSName "my-aks" `
      -ContainerRegistry Acr `
      -ACRName "myacrname" `
      -Location "westus2" `
      -VaultUrl "https://<your-vault>.vault.azure.net" `
      -VaultClientId "<client-id>" `
      -AzureApiClientId "<api-client-id>" `
      -VaultClientSecret "<client-secret>" `
      -VaultTenantId "<tenant-id>" `
      -IngressHost "influxion.example.com" `
      -AcmeEmail "you@example.com"
    

GitHub Container Registry

Use GitHub Container Registry when images are published as ghcr.io/nedomkull-mathematical-modeling/influxion-app:<tag> and ghcr.io/nedomkull-mathematical-modeling/influxion-worker:<tag>. The included GitHub Actions workflow publishes both images to GHCR on pushes to main and on v* tags.

If the packages are public, AKS can pull them without a secret. If they are private, provide -GitHubUsername and -GitHubToken so the script can create the ghcr-pull-secret image pull secret in the influxion namespace.

$env:GHCR_TOKEN = "<github-token-with-read:packages-and-write:packages>"

.\deploy_to_aks.ps1 `
  -ResourceGroupName "my-rg" `
  -AKSName "my-aks" `
  -ContainerRegistry Github `
  -GitHubOwner "nedomkull-mathematical-modeling" `
  -GitHubUsername "<github-username>" `
  -Location "westus2" `
  -VaultUrl "https://<your-vault>.vault.azure.net" `
  -VaultClientId "<client-id>" `
  -AzureApiClientId "<api-client-id>" `
  -VaultClientSecret "<client-secret>" `
  -VaultTenantId "<tenant-id>" `
  -IngressHost "influxion.example.com" `
  -AcmeEmail "you@example.com"

Use -SkipImageBuildPush when the GHCR images have already been built and published by GitHub Actions.

The Helm chart keeps the API and Celery workers in the same pod so generated PDFs can remain short-lived in pod-local runtime storage. Keep app.replicaCount at 1 unless the PDF runtime storage strategy changes.