agent-framework/dotnet/samples/GettingStarted/Workflows/Concurrent/Concurrent/Program.cs

// Copyright (c) Microsoft. All rights reserved.

using Azure.AI.OpenAI;
using Azure.Identity;
using Microsoft.Agents.AI;
using Microsoft.Agents.AI.Workflows;
using Microsoft.Extensions.AI;

namespace WorkflowConcurrentSample;

/// <summary>
/// This sample introduces concurrent execution using "fan-out" and "fan-in" patterns.
///
/// Unlike sequential workflows where executors run one after another, this workflow
/// runs multiple executors in parallel to process the same input simultaneously.
///
/// The workflow structure:
/// 1. StartExecutor sends the same question to two AI agents concurrently (fan-out)
/// 2. Physicist Agent and Chemist Agent answer independently and in parallel
/// 3. AggregationExecutor collects both responses and combines them (fan-in)
///
/// This pattern is useful when you want multiple perspectives on the same input,
/// or when you can break work into independent parallel tasks for better performance.
/// </summary>
/// <remarks>
/// Pre-requisites:
/// - Foundational samples should be completed first.
/// - An Azure OpenAI chat completion deployment must be configured.
/// </remarks>
public static class Program
{
    private static async Task Main()
    {
        // Set up the Azure OpenAI client
        var endpoint = Environment.GetEnvironmentVariable("AZURE_OPENAI_ENDPOINT") ?? throw new InvalidOperationException("AZURE_OPENAI_ENDPOINT is not set.");
        var deploymentName = Environment.GetEnvironmentVariable("AZURE_OPENAI_DEPLOYMENT_NAME") ?? "gpt-4o-mini";
        var chatClient = new AzureOpenAIClient(new Uri(endpoint), new AzureCliCredential()).GetChatClient(deploymentName).AsIChatClient();

        // Create the executors
        ChatClientAgent physicist = new(
            chatClient,
            name: "Physicist",
            instructions: "You are an expert in physics. You answer questions from a physics perspective."
        );
        ChatClientAgent chemist = new(
            chatClient,
            name: "Chemist",
            instructions: "You are an expert in chemistry. You answer questions from a chemistry perspective."
        );
        var startExecutor = new ConcurrentStartExecutor();
        var aggregationExecutor = new ConcurrentAggregationExecutor();

        // Build the workflow by adding executors and connecting them
        var workflow = new WorkflowBuilder(startExecutor)
            .AddFanOutEdge(startExecutor, [physicist, chemist])
            .AddFanInBarrierEdge([physicist, chemist], aggregationExecutor)
            .WithOutputFrom(aggregationExecutor)
            .Build();

        // Execute the workflow in streaming mode
        await using StreamingRun run = await InProcessExecution.RunStreamingAsync(workflow, input: "What is temperature?");
        await foreach (WorkflowEvent evt in run.WatchStreamAsync())
        {
            if (evt is WorkflowOutputEvent output)
            {
                Console.WriteLine($"Workflow completed with results:\n{output.Data}");
            }
        }
    }
}

/// <summary>
/// Executor that starts the concurrent processing by sending messages to the agents.
/// </summary>
internal sealed partial class ConcurrentStartExecutor() :
    Executor("ConcurrentStartExecutor")
{
    /// <summary>
    /// Starts the concurrent processing by sending messages to the agents.
    /// </summary>
    /// <param name="message">The user message to process</param>
    /// <param name="context">Workflow context for accessing workflow services and adding events</param>
    /// <param name="cancellationToken">The <see cref="CancellationToken"/> to monitor for cancellation requests.
    /// The default is <see cref="CancellationToken.None"/>.</param>
    /// <returns>A task representing the asynchronous operation</returns>
    [MessageHandler]
    public async ValueTask HandleAsync(string message, IWorkflowContext context, CancellationToken cancellationToken = default)
    {
        // Broadcast the message to all connected agents. Receiving agents will queue
        // the message but will not start processing until they receive a turn token.
        await context.SendMessageAsync(new ChatMessage(ChatRole.User, message), cancellationToken: cancellationToken);
        // Broadcast the turn token to kick off the agents.
        await context.SendMessageAsync(new TurnToken(emitEvents: true), cancellationToken: cancellationToken);
    }
}

/// <summary>
/// Executor that aggregates the results from the concurrent agents.
/// </summary>
internal sealed class ConcurrentAggregationExecutor() :
    Executor<List<ChatMessage>>("ConcurrentAggregationExecutor")
{
    private readonly List<ChatMessage> _messages = [];

    /// <summary>
    /// Handles incoming messages from the agents and aggregates their responses.
    /// </summary>
    /// <param name="message">The messages from the agent</param>
    /// <param name="context">Workflow context for accessing workflow services and adding events</param>
    /// <param name="cancellationToken">The <see cref="CancellationToken"/> to monitor for cancellation requests.
    /// The default is <see cref="CancellationToken.None"/>.</param>
    /// <returns>A task representing the asynchronous operation</returns>
    public override async ValueTask HandleAsync(List<ChatMessage> message, IWorkflowContext context, CancellationToken cancellationToken = default)
    {
        this._messages.AddRange(message);

        if (this._messages.Count == 2)
        {
            var formattedMessages = string.Join(Environment.NewLine, this._messages.Select(m => $"{m.AuthorName}: {m.Text}"));
            await context.YieldOutputAsync(formattedMessages, cancellationToken);
        }
    }
}