.Net: ONNX 0.5.0 Add Ogahandle resource managment to Service [WIP]

dotnet/samples/Concepts/ChatCompletion/Onnx_ChatCompletion.cs

@@ -34,7 +34,7 @@ public async Task ServicePromptAsync()
 
         Console.WriteLine("======== Onnx - Chat Completion ========");
 
-        var chatService = new OnnxRuntimeGenAIChatCompletionService(
+        using var chatService = new OnnxRuntimeGenAIChatCompletionService(
             modelId: TestConfiguration.Onnx.ModelId,
             modelPath: TestConfiguration.Onnx.ModelPath);
 
@@ -105,5 +105,7 @@ public async Task ChatPromptAsync()
         reply = await kernel.InvokePromptAsync(chatPrompt.ToString());
 
         Console.WriteLine(reply);
+
+        (kernel.GetRequiredService<IChatCompletionService>() as IDisposable)?.Dispose();
     }
 }

dotnet/samples/Concepts/ChatCompletion/Onnx_ChatCompletionStreaming.cs

@@ -86,6 +86,8 @@ public async Task StreamChatPromptAsync()
         reply = await StreamMessageOutputFromKernelAsync(kernel, chatPrompt.ToString());
 
         Console.WriteLine(reply);
+
+        (kernel.GetRequiredService<IChatCompletionService>() as IDisposable)?.Dispose();
     }
 
     /// <summary>
@@ -135,7 +137,7 @@ public async Task StreamTextFromChatAsync()
         }
     }
 
-    private async Task StartStreamingChatAsync(IChatCompletionService chatCompletionService)
+    private async Task StartStreamingChatAsync(OnnxRuntimeGenAIChatCompletionService chatCompletionService)
     {
         Console.WriteLine("Chat content:");
         Console.WriteLine("------------------------");
@@ -156,9 +158,11 @@ private async Task StartStreamingChatAsync(IChatCompletionService chatCompletion
 
         // Second assistant message
         await StreamMessageOutputAsync(chatCompletionService, chatHistory, AuthorRole.Assistant);
+
+        chatCompletionService.Dispose();
     }
 
-    private async Task StreamMessageOutputAsync(IChatCompletionService chatCompletionService, ChatHistory chatHistory, AuthorRole authorRole)
+    private async Task StreamMessageOutputAsync(OnnxRuntimeGenAIChatCompletionService chatCompletionService, ChatHistory chatHistory, AuthorRole authorRole)
     {
         bool roleWritten = false;
         string fullMessage = string.Empty;

dotnet/samples/Demos/OnnxSimpleRAG/Program.cs

@@ -12,6 +12,7 @@
 using Microsoft.SemanticKernel;
 using Microsoft.SemanticKernel.ChatCompletion;
 using Microsoft.SemanticKernel.Connectors.InMemory;
+using Microsoft.SemanticKernel.Connectors.Onnx;
 using Microsoft.SemanticKernel.Data;
 using Microsoft.SemanticKernel.Embeddings;
 using Microsoft.SemanticKernel.PromptTemplates.Handlebars;
@@ -38,7 +39,7 @@
 var kernel = builder.Build();
 
 // Get the instances of the services
-var chatService = kernel.GetRequiredService<IChatCompletionService>();
+using var chatService = kernel.GetRequiredService<IChatCompletionService>() as OnnxRuntimeGenAIChatCompletionService;
 var embeddingService = kernel.GetRequiredService<ITextEmbeddingGenerationService>();
 
 // Create a vector store and a collection to store information

dotnet/src/Connectors/Connectors.Onnx/OnnxRuntimeGenAIChatCompletionService.cs

@@ -25,6 +25,7 @@ public sealed class OnnxRuntimeGenAIChatCompletionService : IChatCompletionServi
     private readonly JsonSerializerOptions? _jsonSerializerOptions;
     private Model? _model;
     private Tokenizer? _tokenizer;
+    private readonly OgaHandle _ogaHandle = new();
 
     private Dictionary<string, object?> AttributesInternal { get; } = new();
 
@@ -212,5 +213,6 @@ public void Dispose()
     {
         this._tokenizer?.Dispose();
         this._model?.Dispose();
+        this._ogaHandle.Dispose();
     }
 }