Deployment Guide

Deployment Guide Skill

Use this skill to orient users to the azd deployment setup, the hook flow, and how to validate regional/model availability.

Key Files & Locations

• azure.yaml — azd config, services, and hooks (preprovision/postprovision/postdown)
• devops/scripts/azd/preprovision.sh — preflight checks, remote state setup, tfvars/provider config
• devops/scripts/azd/postprovision.sh — post-provision tasks (App Config, local env, etc.)
• devops/scripts/azd/postdown.sh — post-azd down cleanup (optional state delete + purge reminder)
• devops/scripts/azd/helpers/preflight-checks.sh — region + quota checks
• infra/terraform/ — Terraform modules + params
• infra/terraform/params/main.tfvars.*.json — environment-specific TF vars (including model_deployments)

Logical Flow (azd)

1. azd up
2. preprovision hook:
   • Runs preflight checks (tools/auth/providers/region/quota)
   • Resolves AZURE_LOCATION
   • Configures Terraform backend (remote state or local)
   • Generates main.tfvars.json and provider config
3. Terraform apply (infra create)
4. postprovision hook:
   • App Config updates
   • Local .env.local generation
   • Data provisioning / phone config (interactive)
5. azd deploy (service build/push/deploy per azure.yaml)
6. azd down → postdown hook (optional remote state cleanup + purge reminder)

Troubleshooting & Docs

Direct users to:

• docs/getting-started/quickstart.md — first-time azd up
• docs/deployment/README.md — advanced deployment and flags
• docs/operations/troubleshooting.md and TROUBLESHOOTING.md — common errors
• docs/getting-started/README.md — region availability matrix

Regional Availability & Model Checks

Use these Azure CLI commands to validate region/model availability:

# List regions
az account list-locations -o table

# Check Azure OpenAI service availability in a region
az cognitiveservices account list-skus --kind OpenAI --location <region> -o table

# Check Cognitive Services availability (Speech, etc.)
az cognitiveservices account list-skus --kind SpeechServices --location <region> -o table

# Check OpenAI quota usage for a region (requires jq)
az cognitiveservices usage list -l <region> -o json \
  | jq -r '.[] | select(.name.value | startswith("OpenAI.")) | "\(.name.value)\t\(.currentValue)/\(.limit)"'

# List deployments in a specific Azure OpenAI resource
az cognitiveservices account deployment list -g <resource-group> -n <openai-account> -o table

Also point to devops/scripts/azd/helpers/preflight-checks.sh for the exact checks used by azd.

Custom Model Deployment Overrides

There are two layers to adjust:

1. Provisioning (what gets created): edit model_deployments in:

   • infra/terraform/params/main.tfvars.*.json (env-specific)
   • or infra/terraform/terraform.tfvars.example (reference)

2. Runtime selection (what the app uses): set the deployment name:

   • AZURE_OPENAI_CHAT_DEPLOYMENT_ID (preferred for app runtime)
   • App Config key: azure/openai/deployment-id

Example:

azd env set AZURE_OPENAI_CHAT_DEPLOYMENT_ID "<your-deployment-name>"
azd deploy

For agent-specific overrides, update deployment_id in: apps/artagent/backend/registries/agentstore/*/agent.yaml.