azure-spring-cloud-training/13-build-a-spring-ai-microservice-with-azure-openai

13 - Build a Spring AI microservice with Azure OpenAI

Prepare Azure OpenAI Service

First, define the following environment variables for use in creating the AI resources we'll use in this example:

export LOCATION="eastus"
export OPENAI_RESOURCE_NAME="<Azure-OpenAI-resource-name>"

Next, run the following command to create an Azure OpenAI resource in the the default resource group:

az cognitiveservices account create \
    -n ${OPENAI_RESOURCE_NAME} \
    -l ${LOCATION} \
    --kind OpenAI \
    --sku s0 \
    --custom-domain ${OPENAI_RESOURCE_NAME}   

To create the model deployments for text-embedding-ada-002 and gpt-35-turbo-16k in your Azure OpenAI service, execute the following commands:

az cognitiveservices account deployment create \
    -n ${OPENAI_RESOURCE_NAME} \
    --deployment-name text-embedding-ada-002 \
    --model-name text-embedding-ada-002 \
    --model-version "2"  \
    --model-format OpenAI

az cognitiveservices account deployment create \
    -n ${OPENAI_RESOURCE_NAME} \
    --deployment-name gpt-35-turbo-16k \
    --model-name gpt-35-turbo-16k \
    --model-version "0613"  \
    --model-format OpenAI \
    --sku Standard \
    --capacity 120

You'll need to retrieve the endpoint and API key for your Azure OpenAI resource to proceed. To do so, follow the guidance in this doc.

Create a Spring AI microservice

The microservice that we create in this guide is available here.

To create our microservice, we will invoke the Spring Initalizr service from the command line:

NOTE: For now, please use the project link above.

Run locally

First, export the environment variables required by Spring AI:

export SPRING_AI_AZURE_OPENAI_API_KEY=<INSERT_KEY_HERE>
export SPRING_AI_AZURE_OPENAI_ENDPOINT=<INSERT_ENDPOINT_URL_HERE>

Next, start the service by running ./mvnw spring-boot:run.

Finally, open http://localhost:8080 in a browser to access the application.

Deploy to Azure Spring Apps

Use the following command to specify the app name on Azure Spring Apps and to allocate the required resources:

az spring app create \
    --name ai-weather-service \
    --cpu 2 \
    --memory 4Gi \
    --min-replicas 2 \
    --max-replicas 2 \
    --assign-endpoint true

Build the jar package using the command: ./mvnw clean package.

Use the following command to deploy the .jar file for the app:

az spring app deploy \
    --name ai-weather-service \
    --artifact-path target/demo-0.0.1-SNAPSHOT.jar \
    --env SPRING_AI_AZURE_OPENAI_API_KEY=${SPRING_AI_AZURE_OPENAI_API_KEY} SPRING_AI_AZURE_OPENAI_ENDPOINT=${SPRING_AI_AZURE_OPENAI_ENDPOINT} \
    --runtime-version Java_17

Test the project in the cloud

• Go to "Apps" in your Azure Spring Apps instance.
  • Verify that ai-weather-service has a Registration status which says 1/1. This shows that it is correctly registered in Spring Cloud Service Registry.
  • Select ai-weather-service to see more information about the microservice.
• Click the "URL" that is provided to open it in another browser tab, then type Basel into the Question prompt box. Click the Ask AI button to see the results!