El Bruno

Disclaimer: This post was created with the help of AI tools to speed up drafting and formatting.

Hi!

Just in case you want to avoid reading, here is a 8-min video overview.

If you are still here, I’ll (try to) walk you through how I updated the eShopLite scenario to take advantage of .NET Aspire 9.4 and the latest GPT-5 models.
This is a great example of how a simple model upgrade can be the first step — and then how new Aspire capabilities like Azure AI Foundry integration can make your code cleaner, easier to maintain, and more deployment-friendly.

---

Details

Step 1 — Updating the Model

The first thing I did was replace the existing model definition in the eShopLite solution:

• From: gpt-4.1-mini
• To: gpt-5-mini

This is just a small change in the model creation definition, but it gives you access to the latest and fastest GPT-5 models.

📦 Locally, this worked perfectly.
☁️ In Azure, after generating the infrastructure files using:

bashCopyEditazd infra gen

…the deployment went smoothly.

---

Step 2 — Leveraging Aspire 9.4 Azure AI Foundry Integration

Aspire 9.4 introduces built-in support for Azure AI Foundry elements. Instead of manually managing your model references and endpoints, you can now:

• Automatically bind to Azure AI Foundry resources in your Aspire app
• Use cleaner, more readable code
• Simplify deployment to Azure with Foundry-focused Bicep files

After refactoring the code to use this new integration, I ran it locally — and everything worked as expected. Even better, the Azure deployment was more transparent and maintainable.

Here is a sample of the AppHost code

    var chatDeploymentName = "gpt-5-mini";
    var embeddingsDeploymentName = "text-embedding-ada-002";

    // Add Azure AI Foundry project
    var foundry = builder.AddAzureAIFoundry("foundry");

    // Add specific model deployments
    var gpt5mini = foundry.AddDeployment(chatDeploymentName, chatDeploymentName, "2025-08-07", "OpenAI");
    var embeddingsDeployment = foundry.AddDeployment(embeddingsDeploymentName, embeddingsDeploymentName, "2", "OpenAI");

And this is the client consumption in the Products project.

// Add Foundry elements
var azureOpenAiClientName = "foundry";
var chatDeploymentName = builder.Configuration["AI_ChatDeploymentName"] ?? "gpt-5-mini";
var embeddingsDeploymentName = builder.Configuration["AI_embeddingsDeploymentName"] ?? "text-embedding-ada-002";

var client = builder.AddOpenAIClient(azureOpenAiClientName);
client.AddChatClient(chatDeploymentName);
client.AddEmbeddingGenerator(embeddingsDeploymentName);

---

Step 3 — Deployment with Bicep Files

With the updated approach, the generated Bicep files are clearer:

• Explicitly define Azure AI Foundry resources
• Clearly map Aspire services to those resources
• Reduce deployment complexity

Once deployed, the application was running on GPT-5 with full Aspire 9.4 + Foundry support.

---

Conclusion

This upgrade journey shows how Aspire 9.4 and Azure AI Foundry integration can significantly improve both the development and deployment experience.
With just a few steps, we moved from a basic GPT-4.1-mini setup to a cleaner, faster, and smarter GPT-5 architecture, ready to run locally or in Azure.

If you’re already using Aspire, upgrading to 9.4 and leveraging Foundry is a no-brainer.

---

Resources

• 📂 eShopLite scenarios: https://aka.ms/eshoplite/repo
• 📚 Aspire 9.4 Azure AI Foundry integration: https://learn.microsoft.com/en-us/dotnet/aspire/whats-new/dotnet-aspire-9.4#-azure-ai-foundry-integration

---