[.Net] add ollama-sample and adds more tests

dotnet/AutoGen.sln

@@ -43,6 +43,8 @@ Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "AutoGen.Ollama", "src\AutoG
 EndProject
 Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "AutoGen.Ollama.Tests", "test\AutoGen.Ollama.Tests\AutoGen.Ollama.Tests.csproj", "{03E31CAA-3728-48D3-B936-9F11CF6C18FE}"
 EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "AutoGen.Ollama.Sample", "sample\AutoGen.Ollama.Sample\AutoGen.Ollama.Sample.csproj", "{93AA4D0D-6EE4-44D5-AD77-7F73A3934544}"
+EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
 		Debug|Any CPU = Debug|Any CPU
@@ -117,6 +119,10 @@ Global
 		{03E31CAA-3728-48D3-B936-9F11CF6C18FE}.Debug|Any CPU.Build.0 = Debug|Any CPU
 		{03E31CAA-3728-48D3-B936-9F11CF6C18FE}.Release|Any CPU.ActiveCfg = Release|Any CPU
 		{03E31CAA-3728-48D3-B936-9F11CF6C18FE}.Release|Any CPU.Build.0 = Release|Any CPU
+		{93AA4D0D-6EE4-44D5-AD77-7F73A3934544}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{93AA4D0D-6EE4-44D5-AD77-7F73A3934544}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{93AA4D0D-6EE4-44D5-AD77-7F73A3934544}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{93AA4D0D-6EE4-44D5-AD77-7F73A3934544}.Release|Any CPU.Build.0 = Release|Any CPU
 	EndGlobalSection
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE
@@ -139,6 +145,7 @@ Global
 		{B61388CA-DC73-4B7F-A7B2-7B9A86C9229E} = {F823671B-3ECA-4AE6-86DA-25E920D3FE64}
 		{9F9E6DED-3D92-4970-909A-70FC11F1A665} = {18BF8DD7-0585-48BF-8F97-AD333080CE06}
 		{03E31CAA-3728-48D3-B936-9F11CF6C18FE} = {F823671B-3ECA-4AE6-86DA-25E920D3FE64}
+		{93AA4D0D-6EE4-44D5-AD77-7F73A3934544} = {FBFEAD1F-29EB-4D99-A672-0CD8473E10B9}
 	EndGlobalSection
 	GlobalSection(ExtensibilityGlobals) = postSolution
 		SolutionGuid = {93384647-528D-46C8-922C-8DB36A382F0B}

dotnet/sample/AutoGen.Ollama.Sample/AutoGen.Ollama.Sample.csproj

@@ -0,0 +1,25 @@
+<Project Sdk="Microsoft.NET.Sdk">
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+    <TargetFramework>$(TestTargetFramework)</TargetFramework>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <GenerateDocumentationFile>True</GenerateDocumentationFile>
+    <NoWarn>$(NoWarn);CS8981;CS8600;CS8602;CS8604;CS8618;CS0219;SKEXP0054;SKEXP0050;SKEXP0110</NoWarn>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <ProjectReference Include="..\..\src\AutoGen.DotnetInteractive\AutoGen.DotnetInteractive.csproj" />
+    <ProjectReference Include="..\..\src\AutoGen.Ollama\AutoGen.Ollama.csproj" />
+    <ProjectReference Include="..\..\src\AutoGen.SourceGenerator\AutoGen.SourceGenerator.csproj" OutputItemType="Analyzer" ReferenceOutputAssembly="false" />
+    <ProjectReference Include="..\..\src\AutoGen\AutoGen.csproj" />
+    <PackageReference Include="FluentAssertions" Version="$(FluentAssertionVersion)" />
+    <PackageReference Include="Microsoft.SemanticKernel.Plugins.Web" Version="$(SemanticKernelExperimentalVersion)" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <None Update="images\*.png">
+      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
+    </None>
+  </ItemGroup>
+
+</Project>

dotnet/sample/AutoGen.Ollama.Sample/Chat_With_LLaMA.cs

@@ -0,0 +1,28 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Chat_With_LLaMA.cs
+
+using AutoGen.Core;
+using AutoGen.Ollama.Extension;
+
+namespace AutoGen.Ollama.Sample;
+
+public class Chat_With_LLaMA
+{
+    public static async Task RunAsync()
+    {
+        using var httpClient = new HttpClient()
+        {
+            BaseAddress = new Uri("https://2xbvtxd1-11434.usw2.devtunnels.ms")
+        };
+
+        var ollamaAgent = new OllamaAgent(
+            httpClient: httpClient,
+            name: "ollama",
+            modelName: "llama3:latest",
+            systemMessage: "You are a helpful AI assistant")
+            .RegisterMessageConnector()
+            .RegisterPrintMessage();
+
+        var reply = await ollamaAgent.SendAsync("Can you write a piece of C# code to calculate 100th of fibonacci?");
+    }
+}

dotnet/sample/AutoGen.Ollama.Sample/Chat_With_LLaVA.cs

@@ -0,0 +1,40 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Chat_With_LLaVA.cs
+
+using AutoGen.Core;
+using AutoGen.Ollama.Extension;
+
+namespace AutoGen.Ollama.Sample;
+
+public class Chat_With_LLaVA
+{
+    public static async Task RunAsync()
+    {
+        using var httpClient = new HttpClient()
+        {
+            BaseAddress = new Uri("https://2xbvtxd1-11434.usw2.devtunnels.ms")
+        };
+
+        var ollamaAgent = new OllamaAgent(
+            httpClient: httpClient,
+            name: "ollama",
+            modelName: "llava:latest",
+            systemMessage: "You are a helpful AI assistant")
+            .RegisterMessageConnector()
+            .RegisterPrintMessage();
+
+        var image = Path.Combine("images", "background.png");
+        var binaryData = BinaryData.FromBytes(File.ReadAllBytes(image), "image/png");
+        var imageMessage = new ImageMessage(Role.User, binaryData);
+        var textMessage = new TextMessage(Role.User, "what's in this image?");
+        var reply = await ollamaAgent.SendAsync(chatHistory: [textMessage, imageMessage]);
+
+        // You can also use MultiModalMessage to put text and image together in one message
+        // In this case, all the messages in the multi-modal message will be put into single piece of message
+        // where the text is the concatenation of all the text messages seperated by \n
+        // and the images are all the images in the multi-modal message
+        var multiModalMessage = new MultiModalMessage(Role.User, [textMessage, imageMessage]);
+
+        reply = await ollamaAgent.SendAsync(chatHistory: [multiModalMessage]);
+    }
+}

dotnet/sample/AutoGen.Ollama.Sample/Program.cs

@@ -0,0 +1,6 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Program.cs
+
+using AutoGen.Ollama.Sample;
+
+await Chat_With_LLaVA.RunAsync();

dotnet/src/AutoGen.Ollama/Middlewares/OllamaMessageConnector.cs

@@ -64,13 +64,9 @@ public async IAsyncEnumerable<IStreamingMessage> InvokeAsync(MiddlewareContext c
 
         // if the chunks are not empty, aggregate them into a single message
         var messageContent = string.Join(string.Empty, chunks.Select(c => c.Message?.Value));
-        var message = new Message
-        {
-            Role = "assistant",
-            Value = messageContent,
-        };
+        var message = new TextMessage(Role.Assistant, messageContent, agent.Name);
 
-        yield return MessageEnvelope.Create(message, agent.Name);
+        yield return message;
     }
 
     private IEnumerable<IMessage> ProcessMessage(IEnumerable<IMessage> messages, IAgent agent)
@@ -96,18 +92,25 @@ private IEnumerable<IMessage> ProcessMessage(IEnumerable<IMessage> messages, IAg
 
     private IEnumerable<IMessage> ProcessMultiModalMessage(MultiModalMessage multiModalMessage, IAgent agent)
     {
-        var messages = new List<IMessage>();
-        foreach (var message in multiModalMessage.Content)
+        var textMessages = multiModalMessage.Content.Where(m => m is TextMessage textMessage && textMessage.GetContent() is not null);
+        var imageMessages = multiModalMessage.Content.Where(m => m is ImageMessage);
+
+        // aggregate the text messages into one message
+        // by concatenating the content using newline
+        var textContent = string.Join("\n", textMessages.Select(m => ((TextMessage)m).Content));
+
+        // collect all the images
+        var images = imageMessages.SelectMany(m => ProcessImageMessage((ImageMessage)m, agent)
+                    .SelectMany(m => (m as IMessage<Message>)?.Content.Images));
+
+        var message = new Message()
         {
-            messages.AddRange(message switch
-            {
-                TextMessage textMessage => ProcessTextMessage(textMessage, agent),
-                ImageMessage imageMessage => ProcessImageMessage(imageMessage, agent),
-                _ => throw new InvalidOperationException("Invalid message type"),
-            });
-        }
+            Role = "user",
+            Value = textContent,
+            Images = images.ToList(),
+        };
 
-        return messages;
+        return [MessageEnvelope.Create(message, agent.Name)];
     }
 
     private IEnumerable<IMessage> ProcessImageMessage(ImageMessage imageMessage, IAgent agent)

dotnet/test/AutoGen.Ollama.Tests/OllamaMessageTests.cs

@@ -2,13 +2,10 @@
 // OllamaMessageTests.cs
 
 using AutoGen.Core;
-using AutoGen.Ollama;
 using AutoGen.Tests;
 using FluentAssertions;
 using Xunit;
-using Message = AutoGen.Ollama.Message;
-
-namespace Autogen.Ollama.Tests;
+namespace AutoGen.Ollama.Tests;
 
 public class OllamaMessageTests
 {
@@ -42,6 +39,36 @@ public async Task ItProcessUserTextMessageAsync()
         await agent.SendAsync(userMessage);
     }
 
+    [Fact]
+    public async Task ItProcessStreamingTextMessageAsync()
+    {
+        var messageConnector = new OllamaMessageConnector();
+        var agent = new EchoAgent("assistant")
+            .RegisterStreamingMiddleware(messageConnector);
+
+        var messageChunks = Enumerable.Range(0, 10)
+            .Select(i => new ChatResponseUpdate()
+            {
+                Message = new Message()
+                {
+                    Value = i.ToString(),
+                    Role = "assistant",
+                }
+            })
+            .Select(m => MessageEnvelope.Create(m));
+
+        IStreamingMessage? finalReply = null;
+        await foreach (var reply in agent.GenerateStreamingReplyAsync(messageChunks))
+        {
+            reply.Should().BeAssignableTo<IStreamingMessage>();
+            finalReply = reply;
+        }
+
+        finalReply.Should().BeOfType<TextMessage>();
+        var textMessage = (TextMessage)finalReply!;
+        textMessage.GetContent().Should().Be("0123456789");
+    }
+
     [Fact]
     public async Task ItProcessAssistantTextMessageAsync()
     {
@@ -126,17 +153,13 @@ public async Task ItProcessMultiModalMessageAsync()
         var agent = new EchoAgent("assistant")
             .RegisterMiddleware(async (msgs, _, innerAgent, ct) =>
             {
-                msgs.Count().Should().Be(2);
-                var textMessage = msgs.First();
-                textMessage.Should().BeOfType<MessageEnvelope<Message>>();
-                var message = (IMessage<Message>)textMessage;
-                message.Content.Role.Should().Be("user");
+                msgs.Count().Should().Be(1);
+                var message = msgs.First();
+                message.Should().BeOfType<MessageEnvelope<Message>>();
 
-                var imageMessage = msgs.Last();
-                imageMessage.Should().BeOfType<MessageEnvelope<Message>>();
-                message = (IMessage<Message>)imageMessage;
-                message.Content.Role.Should().Be("user");
-                message.Content.Images!.Count.Should().Be(1);
+                var multiModalMessage = (IMessage<Message>)message;
+                multiModalMessage.Content.Images!.Count.Should().Be(1);
+                multiModalMessage.Content.Value.Should().Be("Hello");
 
                 return await innerAgent.GenerateReplyAsync(msgs);
             })