Setting Up the Development Environment for Azure AI Foundry

July 2, 2026 · View on GitHub

This guide sets up Azure AI Foundry models for the Java AI apps in this course, using keyless authentication (Microsoft Entra ID) — no API keys to manage. New to the tooling? Start with the development environment guide.

This guide sets up Azure AI Foundry models for the Java AI apps in this course. You have two paths:

  • Option A — Provision with azd + Bicep (recommended): one command deploys the Foundry account and models as code. No portal clicking.
  • Option B — Create resources manually in the Azure AI Foundry portal.

Both paths use keyless authentication (Microsoft Entra ID) — there are no API keys to copy or leak.

Table of Contents

What Gets Created

The Bicep templates in infra/ provision:

  • An Azure AI Foundry account (Microsoft.CognitiveServices/accounts, kind AIServices) with a project
  • A chat deployment — gpt-4o-mini
  • An embedding deployment — text-embedding-3-small (used in later chapters)
  • A keyless role assignment (Cognitive Services OpenAI User) so you sign in with az login instead of managing keys

Prerequisites

From the 02-SetupDevEnvironment folder:

cd 02-SetupDevEnvironment

# Sign in (both tools)
azd auth login
az login

# Provision the Foundry account + model deployments
azd up

azd prompts for an environment name (for example genai-java) and a region. Choose a region where gpt-4o-mini and text-embedding-3-small are available — for example eastus2 or swedencentral.

When provisioning finishes, azd:

  1. Deploys everything defined in infra/main.bicep.
  2. Runs a postprovision hook that writes examples/basic-chat-azure/.env with your endpoint and deployment names (no secrets).

Tip: Re-run azd up any time to apply changes. Run azd down to delete everything and stop incurring cost.

To see the generated settings:

azd env get-values

Now skip to Test Your Setup.

Option B: Create Resources Manually

Prefer the portal? Create the resources by hand:

  1. Go to the Azure AI Foundry portal and sign in.
  2. Create a project (this also creates an AI Foundry resource). Give it a name like GenAIJava.
  3. In your project, open Models + endpoints → Deploy model → Deploy base model.
  4. Deploy gpt-4o-mini (deployment name gpt-4o-mini). Repeat for text-embedding-3-small if you want the embedding examples.
  5. From Overview, copy the endpoint (for example https://<resource>.openai.azure.com/).
  6. Grant yourself keyless access: on the resource, open Access control (IAM) → Add role assignment → assign Cognitive Services OpenAI User to your account.

Still having trouble? See the Azure AI Foundry documentation.

Configure Your Environment

If you used Option A (azd up), your settings file is already written — there's nothing to configure. Skip to Test Your Setup.

If you used Option B (manual), create the example's .env file yourself:

cd 02-SetupDevEnvironment/examples/basic-chat-azure
cp .env.example .env

Edit .env with your endpoint (no key — auth is keyless):

AZURE_OPENAI_ENDPOINT=https://<your-resource>.openai.azure.com/
AZURE_OPENAI_DEPLOYMENT=gpt-4o-mini

Security note: There is no API key to store. You authenticate with Microsoft Entra ID via az login (locally) or a managed identity (in Azure). The .env file holds only non-secret settings and is already covered by .gitignore.

Test Your Setup

Make sure you're signed in so keyless auth can get a token, then run the example:

cd 02-SetupDevEnvironment/examples/basic-chat-azure

az login          # if you aren't already signed in
mvn clean spring-boot:run

You should see a response from the gpt-4o-mini model!

VS Code users: Press F5 to run. The app loads your .env automatically.

Full example: See the Basic Chat with Azure AI Foundry example for details and troubleshooting.

What's Next?

Setup complete! You now have:

  • Azure AI Foundry with gpt-4o-mini and text-embedding-3-small deployed
  • Keyless authentication (Microsoft Entra ID) — no keys to manage
  • A local .env with your endpoint and deployment names
  • A Java development environment ready to go

Continue to Chapter 3: Core Generative AI Techniques to start building AI applications!

Resources

Additional Resources