Dotnet: Add modelServiceId support to SemanticKernelAgent (#5422)

The `SemanticKernelAgent` class has been updated to include an optional
`modelServiceId` parameter, allowing the specification of a service ID
for the model.

## Why are these changes needed?

Currently, `SemanticKernelAgent` uses the parameterless method for
resolving `IChatCompletionSerivce`. This will fail, when multiple models
are registered in the Kernel.

To support different models registered in the Kernel, I adopted the
resolving of the `IChatCompletionSerivce` within the
`SemanticKernelAgent` with an optional parameter. When it is not set, I
resolve the default instance, otherwise, I use the optional parameter as
a servide id for resolving the `IChatCompletionSerivce` service.

## Related issue number



## Checks

- [x] I've included any doc changes needed for
https://microsoft.github.io/autogen/. See
https://microsoft.github.io/autogen/docs/Contribute#documentation to
build and test documentation locally.
- [x] I've added tests (if relevant) corresponding to the changes
introduced in this PR.
- [x] I've made sure all auto checks have passed.

---------

Co-authored-by: Ryan Sweet <rysweet@microsoft.com>
Co-authored-by: Xiaoyun Zhang <bigmiao.zhang@gmail.com>

dotnet/samples/AgentChat/AutoGen.Basic.Sample/Example10_SemanticKernel.cs

@@ -50,7 +50,7 @@ public static async Task RunAsync()
 
         kernel.Plugins.AddFromObject(new LightPlugin());
         var skAgent = kernel
-            .ToSemanticKernelAgent(name: "assistant", systemMessage: "You control the light", settings);
+            .ToSemanticKernelAgent(name: "assistant", systemMessage: "You control the light", settings: settings);
 
         // Send a message to the skAgent, the skAgent supports the following message types:
         // - IMessage<ChatMessageContent>

dotnet/src/AutoGen.SemanticKernel/Extension/KernelExtension.cs

@@ -8,9 +8,9 @@ namespace AutoGen.SemanticKernel.Extension;
 
 public static class KernelExtension
 {
-    public static SemanticKernelAgent ToSemanticKernelAgent(this Kernel kernel, string name, string systemMessage = "You are a helpful AI assistant", PromptExecutionSettings? settings = null)
+    public static SemanticKernelAgent ToSemanticKernelAgent(this Kernel kernel, string name, string systemMessage = "You are a helpful AI assistant", string? modelServiceId = null, PromptExecutionSettings? settings = null)
     {
-        return new SemanticKernelAgent(kernel, name, systemMessage, settings);
+        return new SemanticKernelAgent(kernel, name, systemMessage, modelServiceId, settings);
     }
 
     /// <summary>

dotnet/src/AutoGen.SemanticKernel/SemanticKernelAgent.cs

@@ -32,17 +32,28 @@ public class SemanticKernelAgent : IStreamingAgent
 {
     private readonly Kernel _kernel;
     private readonly string _systemMessage;
+    private readonly string? _modelServiceId;
     private readonly PromptExecutionSettings? _settings;
 
+    /// <summary>
+    /// Create a new instance of <see cref="SemanticKernelAgent"/>
+    /// </summary>
+    /// <param name="kernel">The Semantic Kernel - Kernel object</param>
+    /// <param name="name">The name of the agent.</param>
+    /// <param name="systemMessage">The system message.</param>
+    /// <param name="modelServiceId">Optional serviceId for the model.</param>
+    /// <param name="settings">The prompt execution settings.</param>
     public SemanticKernelAgent(
         Kernel kernel,
         string name,
         string systemMessage = "You are a helpful AI assistant",
+        string? modelServiceId = null,
         PromptExecutionSettings? settings = null)
     {
         _kernel = kernel;
         this.Name = name;
         _systemMessage = systemMessage;
+        _modelServiceId = modelServiceId;
         _settings = settings;
     }
 
@@ -52,7 +63,7 @@ public async Task<IMessage> GenerateReplyAsync(IEnumerable<IMessage> messages, G
     {
         var chatHistory = BuildChatHistory(messages);
         var option = BuildOption(options);
-        var chatService = _kernel.GetRequiredService<IChatCompletionService>();
+        var chatService = GetChatCompletionService();
 
         var reply = await chatService.GetChatMessageContentsAsync(chatHistory, option, _kernel, cancellationToken);
 
@@ -71,7 +82,7 @@ public async IAsyncEnumerable<IMessage> GenerateStreamingReplyAsync(
     {
         var chatHistory = BuildChatHistory(messages);
         var option = BuildOption(options);
-        var chatService = _kernel.GetRequiredService<IChatCompletionService>();
+        var chatService = GetChatCompletionService();
         var response = chatService.GetStreamingChatMessageContentsAsync(chatHistory, option, _kernel, cancellationToken);
 
         await foreach (var content in response)
@@ -108,6 +119,13 @@ private PromptExecutionSettings BuildOption(GenerateReplyOptions? options)
         };
     }
 
+    private IChatCompletionService GetChatCompletionService()
+    {
+        return string.IsNullOrEmpty(_modelServiceId)
+            ? _kernel.GetRequiredService<IChatCompletionService>()
+            : _kernel.GetRequiredService<IChatCompletionService>(_modelServiceId);
+    }
+
     private IEnumerable<ChatMessageContent> ProcessMessage(IEnumerable<IMessage> messages)
     {
         return messages.Select(m => m switch

dotnet/test/AutoGen.SemanticKernel.Tests/SemanticKernelAgentTest.cs

@@ -37,25 +37,25 @@ public async Task BasicConversationTestAsync()
             .AddAzureOpenAIChatCompletion(deploymentName, endpoint, key);
 
         var kernel = builder.Build();
-
-        kernel.GetRequiredService<IChatCompletionService>();
-
         var skAgent = new SemanticKernelAgent(kernel, "assistant");
 
-        var chatMessageContent = MessageEnvelope.Create(new ChatMessageContent(AuthorRole.Assistant, "Hello"));
-        var reply = await skAgent.SendAsync(chatMessageContent);
+        await TestBasicConversationAsync(skAgent);
+    }
 
-        reply.Should().BeOfType<MessageEnvelope<ChatMessageContent>>();
-        reply.As<MessageEnvelope<ChatMessageContent>>().From.Should().Be("assistant");
+    [ApiKeyFact("AZURE_OPENAI_API_KEY", "AZURE_OPENAI_ENDPOINT", "AZURE_OPENAI_DEPLOY_NAME")]
+    public async Task BasicConversationTestWithKeyedServiceAsync()
+    {
+        var endpoint = Environment.GetEnvironmentVariable("AZURE_OPENAI_ENDPOINT") ?? throw new Exception("Please set AZURE_OPENAI_ENDPOINT environment variable.");
+        var key = Environment.GetEnvironmentVariable("AZURE_OPENAI_API_KEY") ?? throw new Exception("Please set AZURE_OPENAI_API_KEY environment variable.");
+        var deploymentName = Environment.GetEnvironmentVariable("AZURE_OPENAI_DEPLOY_NAME") ?? throw new Exception("Please set AZURE_OPENAI_DEPLOY_NAME environment variable.");
+        var modelServiceId = "my-service-id";
+        var builder = Kernel.CreateBuilder()
+            .AddAzureOpenAIChatCompletion(deploymentName, endpoint, key, modelServiceId);
 
-        // test streaming
-        var streamingReply = skAgent.GenerateStreamingReplyAsync(new[] { chatMessageContent });
+        var kernel = builder.Build();
+        var skAgent = new SemanticKernelAgent(kernel, "assistant", modelServiceId: modelServiceId);
 
-        await foreach (var streamingMessage in streamingReply)
-        {
-            streamingMessage.Should().BeOfType<MessageEnvelope<StreamingChatMessageContent>>();
-            streamingMessage.As<MessageEnvelope<StreamingChatMessageContent>>().From.Should().Be("assistant");
-        }
+        await TestBasicConversationAsync(skAgent);
     }
 
     [ApiKeyFact("AZURE_OPENAI_API_KEY", "AZURE_OPENAI_ENDPOINT", "AZURE_OPENAI_DEPLOY_NAME")]
@@ -241,4 +241,22 @@ public async Task SkChatCompletionAgentPluginTestAsync()
         reply.GetContent()!.ToLower().Should().Contain("seattle");
         reply.GetContent()!.ToLower().Should().Contain("sunny");
     }
+
+    private static async Task TestBasicConversationAsync(SemanticKernelAgent agent)
+    {
+        var chatMessageContent = MessageEnvelope.Create(new ChatMessageContent(AuthorRole.Assistant, "Hello"));
+        var reply = await agent.SendAsync(chatMessageContent);
+
+        reply.Should().BeOfType<MessageEnvelope<ChatMessageContent>>();
+        reply.As<MessageEnvelope<ChatMessageContent>>().From.Should().Be("assistant");
+
+        // test streaming
+        var streamingReply = agent.GenerateStreamingReplyAsync(new[] { chatMessageContent });
+
+        await foreach (var streamingMessage in streamingReply)
+        {
+            streamingMessage.Should().BeOfType<MessageEnvelope<StreamingChatMessageContent>>();
+            streamingMessage.As<MessageEnvelope<StreamingChatMessageContent>>().From.Should().Be("assistant");
+        }
+    }
 }