How to use Semantic Kernel agents with Azure AI Foundry connected resource OpenAI models (not base models)

[SamAshmori]

In my Foundry project, I have Azure OpenAI models added as connected resources, not as deployed “base models” in the Foundry Models tab.

When using Foundry base models, I can add them to SK agents via the AddAzureAIInferenceChatCompletion and the Foundry project endpoint (…/models). That works fine.

However, for connected Azure OpenAI resources, Foundry provides the method:

var projectClient = new AIProjectClient(new Uri(projectEndpoint), new DefaultAzureCredential());
var chatClient = projectClient.GetAzureOpenAIChatClient(deploymentName, connectionName, apiVersion);

This returns an Azure.AI.OpenAI.ChatClient that does go through Foundry (so I still get analytics), but I can’t find a straightforward way to register this in SK’s agent architecture as a IChatCompletionService.

Is there a way to have chat completions via Foundry connected resources in Semantic Kernel while keeping analytics?