Skip to content
Justin's Digital Garden

AI-102 Azure AI Development Lab Environment

Lab overview

Section titled “Lab overview”

Try to use free or lowest pricing tier of resources and less extra resources for a development period of 1-2 weeks. All resources are removed after.

Possible resources:

  • Azure AI multi‑service resource at lowest tier
  • Azure OpenAI resource using GPT‑nano/mini models
  • Azure AI Search

Avoid APIM, private networking, fine‑tuning, and resources with fixed monthly hosting.

Example Terraform deployment

Section titled “Example Terraform deployment”

Deployment is not tested, but captures deployment information for use during a first manual creation.

providers.tf
# ---
terraform {
  required_version = ">= 1.5.0"


  required_providers {
    azurerm = {
      source  = "hashicorp/azurerm"
      version = ">= 4.0.0"
    }
    random = {
      source  = "hashicorp/random"
      version = ">= 3.5.0"
    }
  }
}


provider "azurerm" {
  features {}
}


# variables.tf
variable "location" {
  description = "Azure region for all resources."
  type        = string
  # or prefer canadacentral
  default     = "eastus"
}


variable "name_prefix" {
  description = "Short prefix used in resource names (letters/numbers only recommended)."
  type        = string
  default     = "aiwk"
}


variable "tags" {
  description = "Tags applied to all resources."
  type        = map(string)
  default     = {
    purpose   = "ai-week-devtest"
    lifecycle = "ephemeral"
  }
}


variable "create_openai" {
  description = "Whether to create the Azure OpenAI (cognitive account kind=OpenAI)."
  type        = bool
  default     = true
}


variable "create_openai_deployment" {
  description = "Whether to create an Azure OpenAI model deployment (requires model availability/quota)."
  type        = bool
  default     = false
}


variable "openai_deployment_name" {
  description = "Deployment name used within Azure OpenAI."
  type        = string
  default     = "devtest"
}


variable "openai_model_name" {
  description = "Model name for azurerm_cognitive_deployment (varies by region). Example: gpt-4o-mini."
  type        = string
  default     = "gpt-4o-mini"
}


variable "openai_model_version" {
  description = "Optional model version. Leave empty to use default."
  type        = string
  default     = ""
}


variable "openai_deployment_sku" {
  description = "Deployment SKU name for the model deployment. Examples: GlobalStandard, Standard, DataZoneStandard."
  type        = string
  default     = "GlobalStandard"
}


# ---
# main.tf
resource "random_string" "suffix" {
  length  = 6
  upper   = false
  numeric = true
  special = false
}


locals {
  # Keep names short to fit Azure naming limits
  suffix      = random_string.suffix.result
  rg_name     = "${var.name_prefix}-rg-${local.suffix}"
  ai_name     = "${var.name_prefix}aisvc${local.suffix}"
  openai_name = "${var.name_prefix}aoai${local.suffix}"
  search_name = "${var.name_prefix}srch${local.suffix}"
}


resource "azurerm_resource_group" "rg" {
  name     = local.rg_name
  location = var.location
  tags     = var.tags
}


# -----------------------------
# Azure AI multi-service resource
# kind = "AIServices" gives the multi-service “superset” access pattern
# sku_name = "F0" is the free tier SKU for many Azure AI services when available
# -----------------------------
resource "azurerm_cognitive_account" "ai_services" {
  name                = local.ai_name
  location            = azurerm_resource_group.rg.location
  resource_group_name = azurerm_resource_group.rg.name


  kind     = "AIServices"
  sku_name = "F0"


  public_network_access_enabled = true
  local_auth_enabled            = true


  tags = var.tags
}


# -----------------------------
# Azure OpenAI resource
# Note: Azure OpenAI uses azurerm_cognitive_account with kind = "OpenAI"
# sku_name is commonly S0 (standard), as there's no general F0 free tier
# -----------------------------
resource "azurerm_cognitive_account" "openai" {
  count               = var.create_openai ? 1 : 0
  name                = local.openai_name
  location            = azurerm_resource_group.rg.location
  resource_group_name = azurerm_resource_group.rg.name


  kind     = "OpenAI"
  sku_name = "S0"


  # Useful for SDKs that assume https://<subdomain>.openai.azure.com/
  # (Optional, but convenient; changing later forces replacement)
  custom_subdomain_name = lower(local.openai_name)


  public_network_access_enabled = true
  local_auth_enabled            = true


  tags = var.tags
}


# Optional: Azure OpenAI model deployment
# Enable with: create_openai_deployment = true
# If model isn't available in the region or quota isn't enabled, apply will fail.
resource "azurerm_cognitive_deployment" "openai_deployment" {
  count = (var.create_openai && var.create_openai_deployment) ? 1 : 0


  name                 = var.openai_deployment_name
  cognitive_account_id = azurerm_cognitive_account.openai[0].id


  model {
    format  = "OpenAI"
    name    = var.openai_model_name
    version = var.openai_model_version != "" ? var.openai_model_version : null
  }


  sku {
    # Possible values include Standard, DataZoneStandard, GlobalStandard, etc.
    name = var.openai_deployment_sku
  }
}


# -----------------------------
# Azure AI Search (free tier)
# sku = "free" provisions in shared cluster and is ideal for small tests/tutorials.
# -----------------------------
resource "azurerm_search_service" "search" {
  name                = lower(local.search_name)
  resource_group_name = azurerm_resource_group.rg.name
  location            = azurerm_resource_group.rg.location


  sku = "free"


  public_network_access_enabled = true
  local_authentication_enabled  = true


  tags = var.tags
}


# ---
# outputs.tf
output "resource_group_name" {
  value = azurerm_resource_group.rg.name
}


output "ai_services_endpoint" {
  description = "Endpoint for the Azure AI multi-service (AIServices) account."
  value       = azurerm_cognitive_account.ai_services.endpoint
}


output "ai_services_primary_key" {
  description = "Primary key for the Azure AI multi-service account (local auth must be enabled)."
  value       = azurerm_cognitive_account.ai_services.primary_access_key
  sensitive   = true
}


output "search_service_name" {
  value = azurerm_search_service.search.name
}


output "search_service_endpoint" {
  value = "https://${azurerm_search_service.search.name}.search.windows.net"
}


output "search_admin_key_primary" {
  value     = azurerm_search_service.search.primary_key
  sensitive = true
}


output "openai_endpoint" {
  description = "Endpoint for the Azure OpenAI account (if created)."
  value       = var.create_openai ? azurerm_cognitive_account.openai[0].endpoint : null
}


output "openai_primary_key" {
  description = "Primary key for the Azure OpenAI account (if created)."
  value       = var.create_openai ? azurerm_cognitive_account.openai[0].primary_access_key : null
  sensitive   = true
}


output "openai_deployment_name" {
  description = "OpenAI deployment name (if created)."
  value       = (var.create_openai && var.create_openai_deployment) ? azurerm_cognitive_deployment.openai_deployment[0].name : null
}


# ---
# terraform.tfvars.example
location = "eastus"
name_prefix = "aiwk"


tags = {
  purpose   = "ai-week-devtest"
  owner     = "yourname"
  expires   = "2026-02-13"
}


# OpenAI resource on/off
create_openai = true


# Only enable if you know the model is available in the region + you have quota
create_openai_deployment = false
openai_deployment_name   = "devtest"
openai_model_name        = "gpt-4o-mini"
openai_model_version     = ""
openai_deployment_sku    = "GlobalStandard"