chore: import upstream snapshot with attribution

dotnet/samples/02-agents/AgentProviders/foundry/Agent_Step15_ComputerUse/Agent_Step15_ComputerUse.csproj

@@ -0,0 +1,33 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+    <TargetFrameworks>net10.0</TargetFrameworks>
+
+    <Nullable>enable</Nullable>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <NoWarn>$(NoWarn);OPENAICUA001;MEAI001</NoWarn>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="Azure.Identity" />
+    <PackageReference Include="Microsoft.Extensions.Logging.Console" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <ProjectReference Include="..\..\..\..\..\src\Microsoft.Agents.AI.Foundry\Microsoft.Agents.AI.Foundry.csproj" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <None Update="Assets\cua_browser_search.jpg">
+      <CopyToOutputDirectory>Always</CopyToOutputDirectory>
+    </None>
+    <None Update="Assets\cua_search_results.jpg">
+      <CopyToOutputDirectory>Always</CopyToOutputDirectory>
+    </None>
+    <None Update="Assets\cua_search_typed.jpg">
+      <CopyToOutputDirectory>Always</CopyToOutputDirectory>
+    </None>
+  </ItemGroup>
+
+</Project>

dotnet/samples/02-agents/AgentProviders/foundry/Agent_Step15_ComputerUse/ComputerUseUtil.cs

@@ -0,0 +1,93 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+using Microsoft.Extensions.AI;
+using OpenAI.Responses;
+
+namespace Demo.ComputerUse;
+
+/// <summary>
+/// Enum for tracking the state of the simulated web search flow.
+/// </summary>
+internal enum SearchState
+{
+    Initial,        // Browser search page
+    Typed,          // Text entered in search box
+    PressedEnter   // Enter key pressed, transitioning to results
+}
+
+internal static class ComputerUseUtil
+{
+    internal static async Task<Dictionary<string, string>> UploadScreenshotAssetsAsync(IHostedFileClient fileClient)
+    {
+        string assetsDir = Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "Assets");
+
+        (string key, string fileName)[] files =
+        [
+            ("browser_search", "cua_browser_search.jpg"),
+            ("search_typed", "cua_search_typed.jpg"),
+            ("search_results", "cua_search_results.jpg")
+        ];
+
+        Dictionary<string, string> screenshots = [];
+
+        foreach (var (key, fileName) in files)
+        {
+            HostedFileContent result = await fileClient.UploadAsync(
+                Path.Combine(assetsDir, fileName), new HostedFileClientOptions() { Purpose = "assistants" });
+            screenshots[key] = result.FileId;
+        }
+
+        return screenshots;
+    }
+
+    internal static async Task EnsureDeleteScreenshotAssetsAsync(IHostedFileClient fileClient, Dictionary<string, string> screenshots)
+    {
+        foreach (var (_, fileId) in screenshots)
+        {
+            try
+            {
+                await fileClient.DeleteAsync(fileId);
+            }
+            catch
+            {
+            }
+        }
+    }
+
+    /// <summary>
+    /// Simulates executing a computer action by advancing the state
+    /// and returning the screenshot file ID for the new state.
+    /// </summary>
+    internal static async Task<(SearchState State, string FileId)> GetScreenshotAsync(
+        ComputerCallAction action,
+        SearchState currentState,
+        Dictionary<string, string> screenshots)
+    {
+        if (action.Kind == ComputerCallActionKind.Wait)
+        {
+            await Task.Delay(TimeSpan.FromSeconds(5));
+        }
+
+        SearchState nextState = action.Kind switch
+        {
+            ComputerCallActionKind.Click when currentState == SearchState.Typed => SearchState.PressedEnter,
+            ComputerCallActionKind.Type when action.TypeText is not null => SearchState.Typed,
+            ComputerCallActionKind.KeyPress when IsEnterKey(action) => SearchState.PressedEnter,
+            _ => currentState
+        };
+
+        string imageKey = nextState switch
+        {
+            SearchState.PressedEnter => "search_results",
+            SearchState.Typed => "search_typed",
+            _ => "browser_search"
+        };
+
+        return (nextState, screenshots[imageKey]);
+    }
+
+    private static bool IsEnterKey(ComputerCallAction action) =>
+        action.KeyPressKeyCodes is not null &&
+        (action.KeyPressKeyCodes.Contains("Return", StringComparer.OrdinalIgnoreCase) ||
+         action.KeyPressKeyCodes.Contains("Enter", StringComparer.OrdinalIgnoreCase));
+}

dotnet/samples/02-agents/AgentProviders/foundry/Agent_Step15_ComputerUse/Program.cs

@@ -0,0 +1,112 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+// This sample shows how to use the Computer Use tool with AIProjectClient.AsAIAgent(...).
+
+using Azure.AI.Projects;
+using Azure.Identity;
+using Demo.ComputerUse;
+using Microsoft.Agents.AI;
+using Microsoft.Agents.AI.Foundry;
+using Microsoft.Extensions.AI;
+using OpenAI.Responses;
+
+string endpoint = Environment.GetEnvironmentVariable("FOUNDRY_PROJECT_ENDPOINT") ?? throw new InvalidOperationException("FOUNDRY_PROJECT_ENDPOINT is not set.");
+string deploymentName = Environment.GetEnvironmentVariable("AZURE_AI_COMPUTER_USE_DEPLOYMENT_NAME") ?? "computer-use-preview";
+
+// WARNING: DefaultAzureCredential is convenient for development but requires careful consideration in production.
+// In production, consider using a specific credential (e.g., ManagedIdentityCredential) to avoid
+// latency issues, unintended credential probing, and potential security risks from fallback mechanisms.
+AIProjectClient projectClient = new(new Uri(endpoint), new DefaultAzureCredential());
+using IHostedFileClient fileClient = projectClient.GetProjectOpenAIClient().AsIHostedFileClient();
+
+AIAgent agent = projectClient.AsAIAgent(
+    model: deploymentName,
+    name: "ComputerAgent",
+    instructions: "You are a computer automation assistant.",
+    tools: [FoundryAITool.CreateComputerTool(ComputerToolEnvironment.Browser, 1026, 769)]);
+
+Dictionary<string, string> screenshots = [];
+
+try
+{
+    // Upload pre-captured screenshots that simulate browser state transitions.
+    screenshots = await ComputerUseUtil.UploadScreenshotAssetsAsync(fileClient);
+
+    // Enable auto-truncation for the Responses API.
+    ChatClientAgentRunOptions runOptions = new()
+    {
+        ChatOptions = new ChatOptions
+        {
+            RawRepresentationFactory = (_) => new CreateResponseOptions() { TruncationMode = ResponseTruncationMode.Auto },
+        }
+    };
+
+    // Send the initial request with a screenshot of the browser.
+    ChatMessage message = new(ChatRole.User, [
+        new TextContent("Search for 'OpenAI news'. Type it and submit. Once you see results, the task is complete."),
+        new AIContent() { RawRepresentation = ResponseContentPart.CreateInputImagePart(imageFileId: screenshots["browser_search"], imageDetailLevel: ResponseImageDetailLevel.High) }
+    ]);
+
+    Console.WriteLine("Starting computer use session...");
+
+    AgentSession session = await agent.CreateSessionAsync();
+    AgentResponse response = await agent.RunAsync(message, session: session, options: runOptions);
+
+    SearchState currentState = SearchState.Initial;
+
+    for (int i = 0; i < 10; i++)
+    {
+        // Find the next computer call action.
+        ComputerCallResponseItem? computerCall = response.Messages
+            .SelectMany(m => m.Contents)
+            .Select(c => c.RawRepresentation as ComputerCallResponseItem)
+            .FirstOrDefault(item => item is not null);
+
+        if (computerCall is null)
+        {
+            if (currentState == SearchState.PressedEnter)
+            {
+                Console.WriteLine("No more computer actions. Done.");
+                Console.WriteLine(response);
+                break;
+            }
+
+            // Check if the agent is asking for confirmation to proceed, and if so, respond affirmatively.
+            TextContent? textContent = response.Messages
+                .Where(m => m.Role == ChatRole.Assistant)
+                .SelectMany(m => m.Contents.OfType<TextContent>())
+                .FirstOrDefault();
+
+            if (textContent?.Text is { } text && (
+                text.Contains("Would you like me") ||
+                text.Contains("Should I") ||
+                text.Contains("proceed") ||
+                text.Contains('?')))
+            {
+                response = await agent.RunAsync("Please proceed.", session, runOptions);
+                continue;
+            }
+
+            break;
+        }
+
+        Console.WriteLine($"[{i + 1}] Action: {computerCall!.Action.Kind}");
+
+        // Simulate the action and get the resulting screenshot.
+        (currentState, string fileId) = await ComputerUseUtil.GetScreenshotAsync(computerCall.Action, currentState, screenshots);
+
+        // Send the screenshot back as the computer call output.
+        AIContent callOutput = new()
+        {
+            RawRepresentation = new ComputerCallOutputResponseItem(
+                computerCall.CallId,
+                output: ComputerCallOutput.CreateScreenshotOutput(screenshotImageFileId: fileId))
+        };
+
+        response = await agent.RunAsync([new ChatMessage(ChatRole.User, [callOutput])], session: session, options: runOptions);
+    }
+}
+finally
+{
+    await ComputerUseUtil.EnsureDeleteScreenshotAssetsAsync(fileClient, screenshots);
+}

dotnet/samples/02-agents/AgentProviders/foundry/Agent_Step15_ComputerUse/README.md

@@ -0,0 +1,56 @@
+# Computer Use with the Responses API
+
+This sample shows how to use the Computer Use tool with `AIProjectClient.AsAIAgent(...)`.
+
+## What this sample demonstrates
+
+- Using `FoundryAITool.CreateComputerTool()` to add computer use capabilities
+- Processing computer call actions (click, type, key press)
+- Managing the computer use interaction loop with screenshots
+
+For more information, see [Use the computer tool](https://learn.microsoft.com/en-us/azure/foundry/agents/how-to/tools/computer-use?pivots=csharp).
+
+## How the simulation works
+
+In a real computer use scenario, the model controls a virtual keyboard and mouse to interact with a live browser — typing text, clicking buttons, and pressing keys. The host application captures a screenshot after each action and sends it back to the model so it can decide what to do next.
+
+**This sample does not connect to a real browser.** Instead, it intercepts the model's actions and returns pre-captured screenshots as if the actions were actually performed. No real typing, clicking, or key presses happen — the sample fakes the environment so you can explore the computer use protocol without any browser automation setup.
+
+### State transitions
+
+The model receives a screenshot as input, analyzes it, and responds with a computer action as output. The sample maps each action to a new state and returns the corresponding screenshot:
+
+| Step | Model Action    | What Happens                              | Screenshot Sent Back to Model                                |
+|------|-----------------|-------------------------------------------|--------------------------------------------------------------|
+| 1    |                 | Session starts with the user prompt       | `cua_browser_search.jpg` — empty search page                 |
+| 2    | Click           | Model clicks the search box to focus it   | `cua_browser_search.jpg` — same page                         |
+| 3    | Type            | Model types the search query into the box | `cua_search_typed.jpg` — search text visible in the box      |
+| 3a   | *(text response)* | Model may ask for confirmation instead of acting | `cua_search_typed.jpg` — same page |
+| 4    | KeyPress Enter  | Model presses Enter to submit the search  | `cua_search_results.jpg` — search results page               |
+
+### Interaction loop
+
+1. The user prompt and the initial screenshot (`cua_browser_search.jpg` — an empty search page) are sent to the model as input.
+2. The model analyzes the screenshot and responds with a computer action (e.g., click on the search box to focus it, then type search text, then press Enter).
+3. The sample intercepts the action, advances the state, and sends back the next pre-captured screenshot as if the action was performed on a real browser.
+4. Steps 2–3 repeat until the model stops requesting actions or the iteration limit is reached.
+
+## Prerequisites
+
+- .NET 10 SDK or later
+- Microsoft Foundry service endpoint and deployment configured
+- An authenticated Azure identity (for example, sign in with `az login`)
+
+Set the following environment variables:
+
+```powershell
+$env:FOUNDRY_PROJECT_ENDPOINT="https://your-foundry-service.services.ai.azure.com/api/projects/your-foundry-project"
+$env:AZURE_AI_COMPUTER_USE_DEPLOYMENT_NAME="computer-use-preview"
+```
+
+## Run the sample
+
+```powershell
+dotnet run
+```
+