AI Application Demo Using Terrafrom and AKS

This sample demo app consists of a group of containerized microservices that can be easily deployed into an Azure Kubernetes Service (AKS) cluster. This is meant to show a realistic scenario using a polyglot architecture, event-driven design, and common open source back-end services (eg - RabbitMQ, MongoDB). The application also leverages OpenAI's GPT-3 models to generate product descriptions. This can be done using either Azure OpenAI or OpenAI.

This application is inspired by another demo app called Red Dog.

Note: This is not meant to be an example of perfect code to be used in production, but more about showing a realistic application running in AKS.

Architecture

The application has the following services:

Service	Description
makeline-service	This service handles processing orders from the queue and completing them (Golang)
order-service	This service is used for placing orders (Javascript)
product-service	This service is used to perform CRUD operations on products (Rust)
store-front	Web app for customers to place orders (Vue.js)
store-admin	Web app used by store employees to view orders in queue and manage products (Vue.js)
virtual-customer	Simulates order creation on a scheduled basis (Rust)
virtual-worker	Simulates order completion on a scheduled basis (Rust)
ai-service	Optional service for adding generative text and graphics creation (Python)
mongodb	MongoDB instance for persisted data
rabbitmq	RabbitMQ for an order queue

Prerequisites

For this demo, you can either use Azure OpenAI service or OpenAI service. If you plan on using Azure OpenAI service, you need to request access to enable it on your Azure subscription using the Request Access to Azure OpenAI Service form.

If you plan on using OpenAI, sign up on the OpenAI website.

To Deploy

1. Open a cloud Shell from the Azure portal
2. Clone this repo by running this command:

git clone https://github.com/dawright22/azure-ai-demo.git

3. Change into the directory this created
4. Now copy and Run this command:

Terraform init

6. Now copy and Run this command:

Terraform apply

Once the deployment is complete you will see the outputs of the deployment. Then rename [pki_build.tf.second] to [pki_build.tf] and run terraform apply again. This will deploy the PKI components.

Finally, you will need to run the following command to deploy the application: Rename [app_build.tf.third] to [app_build.tf] and run terraform apply again. This will deploy the application components.

That's all Fokes! [to quote bugs bunny]

If you want to customise location or other componets then you can start playing with the Variables file to suit you.

Terraform Outputs give you all the names of resocurces created so you can use them to connect to the resources.

Requirements

Name	Version
helm	~> 2.0.2
local	~>2.4.0
random	=3.5.1

Providers

Name	Version
azurerm	n/a
helm	~> 2.0.2
kubernetes	n/a
random	=3.5.1

Modules

No modules.

Resources

Name	Type
azurerm_cognitive_account.tf-ai-demo	resource
azurerm_cognitive_deployment.tf-ai-demo	resource
azurerm_cosmosdb_account.tf-ai-demo	resource
azurerm_cosmosdb_mongo_collection.tf-ai-demo	resource
azurerm_cosmosdb_mongo_database.tf-ai-demo	resource
azurerm_federated_identity_credential.tf-ai-demo	resource
azurerm_kubernetes_cluster.tf-ai-demo	resource
azurerm_resource_group.tf-ai-demo	resource
azurerm_role_assignment.tf-ai-demo_aoai_me	resource
azurerm_role_assignment.tf-ai-demo_aoai_mi	resource
azurerm_servicebus_namespace.tf-ai-demo	resource
azurerm_servicebus_namespace_authorization_rule.tf-ai-demo	resource
azurerm_servicebus_queue.tf-ai-demo	resource
azurerm_servicebus_queue_authorization_rule.tf-ai-demo	resource
azurerm_user_assigned_identity.tf-ai-demo	resource
helm_release.CertManger	resource
helm_release.traefik	resource
helm_release.vault	resource
kubernetes_namespace.CertManager	resource
random_integer.tf-ai-demo	resource
random_pet.tf-ai-demo	resource
azurerm_client_config.current	data source
azurerm_subscription.current	data source

Inputs

Name	Description	Type	Default	Required
ai_location	value of azure region for deploying azure ai service	string	"Australia East"	no
k8s_namespace	value of kubernetes namespace	string	"default"	no
location	n/a	string	"Australia East"	no
openai_model_capacity	value of azure openai model capacity	number	120	no
openai_model_name	value of azure openai model name	string	"gpt-35-turbo"	no
openai_model_version	value of azure openai model version	string	"0613"	no

Outputs

Name	Description
ai_endpoint	n/a
ai_key	n/a
ai_managed_identity_client_id	n/a
ai_model_name	n/a
ai_openai_deployment_name	n/a
aks_name	n/a
db_account_name	n/a
db_key	n/a
db_uri	n/a
k8s_namespace	n/a
rg_name	n/a
sb_listener_key	n/a
sb_listener_username	n/a
sb_namespace_host	n/a
sb_namespace_uri	n/a
sb_sender_key	n/a
sb_sender_username	n/a

Additional Resources

• AKS Documentation. https://learn.microsoft.com/azure/aks
• Kubernetes Learning Path. https://azure.microsoft.com/resources/kubernetes-learning-path