How-To: Coordinate Agent Collaboration using Agent Group Chat (Experimental)

  • Article

Warning

The Semantic Kernel Agent Framework is experimental, still in development and is subject to change.

Overview

In this sample, we will explore how to use Agent Group Chat to coordinate collboration of two different agents working to review and rewrite user provided content. Each agent is assigned a distinct role:

  • Reviewer: Reviews and provides direction to Writer.
  • Writer: Updates user content based on Reviewer input.

The approach will be broken down step-by-step to high-light the key parts of the coding process.

Getting Started

Before proceeding with feature coding, make sure your development environment is fully set up and configured.

Start by creating a Console project. Then, include the following package references to ensure all required dependencies are available.

To add package dependencies from the command-line use the dotnet command:

dotnet add package Azure.Identity
dotnet add package Microsoft.Extensions.Configuration
dotnet add package Microsoft.Extensions.Configuration.Binder
dotnet add package Microsoft.Extensions.Configuration.UserSecrets
dotnet add package Microsoft.Extensions.Configuration.EnvironmentVariables
dotnet add package Microsoft.SemanticKernel.Connectors.AzureOpenAI
dotnet add package Microsoft.SemanticKernel.Agents.Core --prerelease

If managing NuGet packages in Visual Studio, ensure Include prerelease is checked.

The project file (.csproj) should contain the following PackageReference definitions:

  <ItemGroup>
    <PackageReference Include="Azure.Identity" Version="<stable>" />
    <PackageReference Include="Microsoft.Extensions.Configuration" Version="<stable>" />
    <PackageReference Include="Microsoft.Extensions.Configuration.Binder" Version="<stable>" />
    <PackageReference Include="Microsoft.Extensions.Configuration.UserSecrets" Version="<stable>" />
    <PackageReference Include="Microsoft.Extensions.Configuration.EnvironmentVariables" Version="<stable>" />
    <PackageReference Include="Microsoft.SemanticKernel.Agents.Core" Version="<latest>" />
    <PackageReference Include="Microsoft.SemanticKernel.Connectors.AzureOpenAI" Version="<latest>" />
  </ItemGroup>

The Agent Framework is experimental and requires warning suppression. This may addressed in as a property in the project file (.csproj):

  <PropertyGroup>
    <NoWarn>$(NoWarn);CA2007;IDE1006;SKEXP0001;SKEXP0110;OPENAI001</NoWarn>
  </PropertyGroup>

import asyncio
import os
import copy
import pyperclip # Install via pip

from semantic_kernel.agents import AgentGroupChat, ChatCompletionAgent
from semantic_kernel.agents.strategies.selection.kernel_function_selection_strategy import (
    KernelFunctionSelectionStrategy,
)
from semantic_kernel.agents.strategies.termination.kernel_function_termination_strategy import (
    KernelFunctionTerminationStrategy,
)
from semantic_kernel.connectors.ai.open_ai.services.azure_chat_completion import AzureChatCompletion
from semantic_kernel.contents.chat_message_content import ChatMessageContent
from semantic_kernel.contents.utils.author_role import AuthorRole
from semantic_kernel.functions.kernel_function_decorator import kernel_function
from semantic_kernel.functions.kernel_function_from_prompt import KernelFunctionFromPrompt
from semantic_kernel.kernel import Kernel

Agents are currently unavailable in Java.

Configuration

This sample requires configuration setting in order to connect to remote services. You will need to define settings for either Open AI or Azure Open AI.

# Open AI
dotnet user-secrets set "OpenAISettings:ApiKey" "<api-key>"
dotnet user-secrets set "OpenAISettings:ChatModel" "gpt-4o"

# Azure Open AI
dotnet user-secrets set "AzureOpenAISettings:ApiKey" "<api-key>" # Not required if using token-credential
dotnet user-secrets set "AzureOpenAISettings:Endpoint" "<model-endpoint>"
dotnet user-secrets set "AzureOpenAISettings:ChatModelDeployment" "gpt-4o"

The following class is used in all of the Agent examples. Be sure to include it in your project to ensure proper functionality. This class serves as a foundational component for the examples that follow.

using System.Reflection;
using Microsoft.Extensions.Configuration;

namespace AgentsSample;

public class Settings
{
    private readonly IConfigurationRoot configRoot;

    private AzureOpenAISettings azureOpenAI;
    private OpenAISettings openAI;

    public AzureOpenAISettings AzureOpenAI => this.azureOpenAI ??= this.GetSettings<Settings.AzureOpenAISettings>();
    public OpenAISettings OpenAI => this.openAI ??= this.GetSettings<Settings.OpenAISettings>();

    public class OpenAISettings
    {
        public string ChatModel { get; set; } = string.Empty;
        public string ApiKey { get; set; } = string.Empty;
    }

    public class AzureOpenAISettings
    {
        public string ChatModelDeployment { get; set; } = string.Empty;
        public string Endpoint { get; set; } = string.Empty;
        public string ApiKey { get; set; } = string.Empty;
    }

    public TSettings GetSettings<TSettings>() =>
        this.configRoot.GetRequiredSection(typeof(TSettings).Name).Get<TSettings>()!;

    public Settings()
    {
        this.configRoot =
            new ConfigurationBuilder()
                .AddEnvironmentVariables()
                .AddUserSecrets(Assembly.GetExecutingAssembly(), optional: true)
                .Build();
    }
}

The quickest way to get started with the proper configuration to run the sample code is to create a .env file at the root of your project (where your script is run).

Configure the following settings in your .env file for either Azure OpenAI or OpenAI:

AZURE_OPENAI_API_KEY="..."
AZURE_OPENAI_ENDPOINT="https://..."
AZURE_OPENAI_CHAT_DEPLOYMENT_NAME="..."
AZURE_OPENAI_API_VERSION="..."

OPENAI_API_KEY="sk-..."
OPENAI_ORG_ID=""
OPENAI_CHAT_MODEL_ID=""

Once configured, the respective AI service classes will pick up the required variables and use them during instantiation.

Agents are currently unavailable in Java.

Coding

The coding process for this sample involves:

  1. Setup - Initializing settings and the plug-in.
  2. Agent Definition - Create the two Chat Completion Agent instances (Reviewer and Writer).
  3. Chat Definition - Create the Agent Group Chat and associated strategies.
  4. The Chat Loop - Write the loop that drives user / agent interaction.

The full example code is provided in the Final section. Refer to that section for the complete implementation.

Setup

Prior to creating any Chat Completion Agent, the configuration settings, plugins, and Kernel must be initialized.

Instantiate the the Settings class referenced in the previous Configuration section.

Settings settings = new();

Agents are currently unavailable in Java.

Now initialize a Kernel instance with an IChatCompletionService.

IKernelBuilder builder = Kernel.CreateBuilder();

builder.AddAzureOpenAIChatCompletion(
	settings.AzureOpenAI.ChatModelDeployment,
	settings.AzureOpenAI.Endpoint,
	new AzureCliCredential());

Kernel kernel = builder.Build();

Initialize the kernel object:

kernel = Kernel()

Agents are currently unavailable in Java.

Let's also create a second Kernel instance via cloning and add a plug-in that will allow the reivew to place updated content on the clip-board.

Kernel toolKernel = kernel.Clone();
toolKernel.Plugins.AddFromType<ClipboardAccess>();

tool_kernel = copy.deepcopy(kernel)
tool_kernel.add_plugin(ClipboardAccess(), plugin_name="clipboard")

Agents are currently unavailable in Java.

The Clipboard plugin may be defined as part of the sample.

private sealed class ClipboardAccess
{
    [KernelFunction]
    [Description("Copies the provided content to the clipboard.")]
    public static void SetClipboard(string content)
    {
        if (string.IsNullOrWhiteSpace(content))
        {
            return;
        }

        using Process clipProcess = Process.Start(
            new ProcessStartInfo
            {
                FileName = "clip",
                RedirectStandardInput = true,
                UseShellExecute = false,
            });

        clipProcess.StandardInput.Write(content);
        clipProcess.StandardInput.Close();
    }
}

Note: we are leveraging a Python package called pyperclip. Please install is using pip.

class ClipboardAccess:
    @kernel_function
    def set_clipboard(content: str):
        if not content.strip():
            return

        pyperclip.copy(content)

Agents are currently unavailable in Java.

Agent Definition

Let's declare the agent names as const so they might be referenced in Agent Group Chat strategies:

const string ReviewerName = "Reviewer";
const string WriterName = "Writer";

REVIEWER_NAME = "Reviewer"
WRITER_NAME = "Writer"

Agents are currently unavailable in Java.

Defining the Reviewer agent uses the pattern explored in How-To: Chat Completion Agent.

Here the Reviewer is given the role of responding to user input, providing direction to the Writer agent, and verifying result of the Writer agent.

ChatCompletionAgent agentReviewer =
    new()
    {
        Name = ReviewerName,
        Instructions =
            """
            Your responsiblity is to review and identify how to improve user provided content.
            If the user has providing input or direction for content already provided, specify how to address this input.
            Never directly perform the correction or provide example.
            Once the content has been updated in a subsequent response, you will review the content again until satisfactory.
            Always copy satisfactory content to the clipboard using available tools and inform user.

            RULES:
            - Only identify suggestions that are specific and actionable.
            - Verify previous suggestions have been addressed.
            - Never repeat previous suggestions.
            """,
        Kernel = toolKernel,
        Arguments =
            new KernelArguments(
                new AzureOpenAIPromptExecutionSettings() 
                { 
                    FunctionChoiceBehavior = FunctionChoiceBehavior.Auto() 
                })
    };

agent_reviewer = ChatCompletionAgent(
    service_id=REVIEWER_NAME,
    kernel=_create_kernel_with_chat_completion(REVIEWER_NAME),
    name=REVIEWER_NAME,
    instructions="""
        Your responsiblity is to review and identify how to improve user provided content.
        If the user has providing input or direction for content already provided, specify how to 
        address this input.
        Never directly perform the correction or provide example.
        Once the content has been updated in a subsequent response, you will review the content 
        again until satisfactory.
        Always copy satisfactory content to the clipboard using available tools and inform user.

        RULES:
        - Only identify suggestions that are specific and actionable.
        - Verify previous suggestions have been addressed.
        - Never repeat previous suggestions.
        """,
)

Agents are currently unavailable in Java.

The Writer agent is is similiar, but doesn't require the specification of Execution Settings since it isn't configured with a plug-in.

Here the Writer is given a single-purpose task, follow direction and rewrite the content.

ChatCompletionAgent agentWriter =
    new()
    {
        Name = WriterName,
        Instructions =
            """
            Your sole responsiblity is to rewrite content according to review suggestions.

            - Always apply all review direction.
            - Always revise the content in its entirety without explanation.
            - Never address the user.
            """,
        Kernel = kernel,
    };

agent_writer = ChatCompletionAgent(
    service_id=COPYWRITER_NAME,
    kernel=_create_kernel_with_chat_completion(COPYWRITER_NAME),
    name=COPYWRITER_NAME,
    instructions="""
        Your sole responsiblity is to rewrite content according to review suggestions.

        - Always apply all review direction.
        - Always revise the content in its entirety without explanation.
        - Never address the user.
        """,
)

Agents are currently unavailable in Java.

Chat Definition

Defining the Agent Group Chat requires considering the strategies for selecting the Agent turn and determining when to exit the Chat loop. For both of these considerations, we will define a Kernel Prompt Function.

The first to reason over Agent selection:

Using AgentGroupChat.CreatePromptFunctionForStrategy provides a convenient mechanism to avoid HTML encoding the message paramter.

KernelFunction selectionFunction =
    AgentGroupChat.CreatePromptFunctionForStrategy(
        $$$"""
        Examine the provided RESPONSE and choose the next participant.
        State only the name of the chosen participant without explanation.
        Never choose the participant named in the RESPONSE.

        Choose only from these participants:
        - {{{ReviewerName}}}
        - {{{WriterName}}}

        Always follow these rules when choosing the next participant:
        - If RESPONSE is user input, it is {{{ReviewerName}}}'s turn.
        - If RESPONSE is by {{{ReviewerName}}}, it is {{{WriterName}}}'s turn.
        - If RESPONSE is by {{{WriterName}}}, it is {{{ReviewerName}}}'s turn.

        RESPONSE:
        {{$lastmessage}}
        """,
        safeParameterNames: "lastmessage");

selection_function = KernelFunctionFromPrompt(
    function_name="selection",
    prompt=f"""
    Determine which participant takes the next turn in a conversation based on the the most recent participant.
    State only the name of the participant to take the next turn.
    No participant should take more than one turn in a row.

    Choose only from these participants:
    - {REVIEWER_NAME}
    - {COPYWRITER_NAME}

    Always follow these rules when selecting the next participant:
    - After user input, it is {COPYWRITER_NAME}'s turn.
    - After {COPYWRITER_NAME} replies, it is {REVIEWER_NAME}'s turn.
    - After {REVIEWER_NAME} provides feedback, it is {COPYWRITER_NAME}'s turn.

    History:
    {{{{$history}}}}
    """,
)

Agents are currently unavailable in Java.

The second will evaluate when to exit the Chat loop:

const string TerminationToken = "yes";

KernelFunction terminationFunction =
    AgentGroupChat.CreatePromptFunctionForStrategy(
        $$$"""
        Examine the RESPONSE and determine whether the content has been deemed satisfactory.
        If content is satisfactory, respond with a single word without explanation: {{{TerminationToken}}}.
        If specific suggestions are being provided, it is not satisfactory.
        If no correction is suggested, it is satisfactory.

        RESPONSE:
        {{$lastmessage}}
        """,
        safeParameterNames: "lastmessage");

TERMINATION_KEYWORD = "yes"

termination_function = KernelFunctionFromPrompt(
    function_name="termination",
    prompt=f"""
        Examine the RESPONSE and determine whether the content has been deemed satisfactory.
        If content is satisfactory, respond with a single word without explanation: {TERMINATION_KEYWORD}.
        If specific suggestions are being provided, it is not satisfactory.
        If no correction is suggested, it is satisfactory.

        RESPONSE:
        {{{{$history}}}}
        """,
)

Agents are currently unavailable in Java.

Both of these Strategies will only require knowledge of the most recent Chat message. This will reduce token usage and help improve performance:

ChatHistoryTruncationReducer historyReducer = new(1);

**ChatHistoryReducer is coming soon to Python.**

Agents are currently unavailable in Java.

Finally we are ready to bring everything together in our Agent Group Chat definition.

Creating AgentGroupChat involves:

  1. Include both agents in the constructor.
  2. Define a KernelFunctionSelectionStrategy using the previously defined KernelFunction and Kernel instance.
  3. Define a KernelFunctionTerminationStrategy using the previously defined KernelFunction and Kernel instance.

Notice that each strategy is responsible for parsing the KernelFunction result.

AgentGroupChat chat =
    new(agentReviewer, agentWriter)
    {
        ExecutionSettings = new AgentGroupChatSettings
        {
            SelectionStrategy =
                new KernelFunctionSelectionStrategy(selectionFunction, kernel)
                {
                    // Always start with the editor agent.
                    InitialAgent = agentReviewer,
                    // Save tokens by only including the final response
                    HistoryReducer = historyReducer,
                    // The prompt variable name for the history argument.
                    HistoryVariableName = "lastmessage",
                    // Returns the entire result value as a string.
                    ResultParser = (result) => result.GetValue<string>() ?? agentReviewer.Name
                },
            TerminationStrategy =
                new KernelFunctionTerminationStrategy(terminationFunction, kernel)
                {
                    // Only evaluate for editor's response
                    Agents = [agentReviewer],
                    // Save tokens by only including the final response
                    HistoryReducer = historyReducer,
                    // The prompt variable name for the history argument.
                    HistoryVariableName = "lastmessage",
                    // Limit total number of turns
                    MaximumIterations = 12,
                    // Customer result parser to determine if the response is "yes"
                    ResultParser = (result) => result.GetValue<string>()?.Contains(TerminationToken, StringComparison.OrdinalIgnoreCase) ?? false
                }
        }
    };

Console.WriteLine("Ready!");

Creating AgentGroupChat involves:

  1. Include both agents in the constructor.
  2. Define a KernelFunctionSelectionStrategy using the previously defined KernelFunction and Kernel instance.
  3. Define a KernelFunctionTerminationStrategy using the previously defined KernelFunction and Kernel instance.

Notice that each strategy is responsible for parsing the KernelFunction result.

chat = AgentGroupChat(
    agents=[agent_writer, agent_reviewer],
    selection_strategy=KernelFunctionSelectionStrategy(
        function=selection_function,
        kernel=_create_kernel_with_chat_completion("selection"),
        result_parser=lambda result: str(result.value[0]) if result.value is not None else COPYWRITER_NAME,
        agent_variable_name="agents",
        history_variable_name="history",
    ),
    termination_strategy=KernelFunctionTerminationStrategy(
        agents=[agent_reviewer],
        function=termination_function,
        kernel=_create_kernel_with_chat_completion("termination"),
        result_parser=lambda result: TERMINATION_KEYWORD in str(result.value[0]).lower(),
        history_variable_name="history",
        maximum_iterations=10,
    ),
)

Agents are currently unavailable in Java.

The Chat Loop

At last, we are able to coordinate the interaction between the user and the Agent Group Chat. Start by creating creating an empty loop.

Note: Unlike the other examples, no external history or thread is managed. Agent Group Chat manages the conversation history internally.

bool isComplete = false;
do
{

} while (!isComplete);

is_complete: bool = False
while not is_complete:
    # operational logic

Agents are currently unavailable in Java.

Now let's capture user input within the previous loop. In this case:

  • Empty input will be ignored
  • The term EXIT will signal that the conversation is completed
  • The term RESET will clear the Agent Group Chat history
  • Any term starting with @ will be treated as a file-path whose content will be provided as input
  • Valid input will be added to the Agent Group Chaty as a User message.
Console.WriteLine();
Console.Write("> ");
string input = Console.ReadLine();
if (string.IsNullOrWhiteSpace(input))
{
    continue;
}
input = input.Trim();
if (input.Equals("EXIT", StringComparison.OrdinalIgnoreCase))
{
    isComplete = true;
    break;
}

if (input.Equals("RESET", StringComparison.OrdinalIgnoreCase))
{
    await chat.ResetAsync();
    Console.WriteLine("[Converation has been reset]");
    continue;
}

if (input.StartsWith("@", StringComparison.Ordinal) && input.Length > 1)
{
    string filePath = input.Substring(1);
    try
    {
        if (!File.Exists(filePath))
        {
            Console.WriteLine($"Unable to access file: {filePath}");
            continue;
        }
        input = File.ReadAllText(filePath);
    }
    catch (Exception)
    {
        Console.WriteLine($"Unable to access file: {filePath}");
        continue;
    }
}

chat.AddChatMessage(new ChatMessageContent(AuthorRole.User, input));

user_input = input("User:> ")
if not user_input:
    continue

if user_input.lower() == "exit":
    is_complete = True
    break

if user_input.lower() == "reset":
    await chat.reset()
    print("[Conversation has been reset]")
    continue

if user_input.startswith("@") and len(input) > 1:
    file_path = input[1:]
    try:
        if not os.path.exists(file_path):
            print(f"Unable to access file: {file_path}")
            continue
        with open(file_path) as file:
            user_input = file.read()
    except Exception:
        print(f"Unable to access file: {file_path}")
        continue

await chat.add_chat_message(ChatMessageContent(role=AuthorRole.USER, content=user_input))

Agents are currently unavailable in Java.

To initate the Agent collaboration in response to user input and display the Agent responses, invoke the Agent Group Chat; however, first be sure to reset the Completion state from any prior invocation.

Note: Service failures are being caught and displayed to avoid crashing the conversation loop.

chat.IsComplete = false;

try
{
    await foreach (ChatMessageContent response in chat.InvokeAsync())
    {
        Console.WriteLine();
        Console.WriteLine($"{response.AuthorName.ToUpperInvariant()}:{Environment.NewLine}{response.Content}");
    }
}
catch (HttpOperationException exception)
{
    Console.WriteLine(exception.Message);
    if (exception.InnerException != null)
    {
        Console.WriteLine(exception.InnerException.Message);
        if (exception.InnerException.Data.Count > 0)
        {
            Console.WriteLine(JsonSerializer.Serialize(exception.InnerException.Data, new JsonSerializerOptions() { WriteIndented = true }));
        }
    }
}

chat.is_complete = False
async for response in chat.invoke():
    print(f"# {response.role} - {response.name or '*'}: '{response.content}'")

if chat.is_complete:
    is_complete = True
    break

Agents are currently unavailable in Java.

Final

Bringing all the steps together, we have the final code for this example. The complete implementation is provided below.

Try using these suggested inputs:

  1. Hi
  2. {"message: "hello world"}
  3. {"message": "hello world"}
  4. Semantic Kernel (SK) is an open-source SDK that enables developers to build and orchestrate complex AI workflows that involve natural language processing (NLP) and machine learning models. It provies a flexible platform for integrating AI capabilities such as semantic search, text summarization, and dialogue systems into applications. With SK, you can easily combine different AI services and models, define thei relationships, and orchestrate interactions between them.
  5. make this two paragraphs
  6. thank you
  7. @.\WomensSuffrage.txt
  8. its good, but is it ready for my college professor?
// Copyright (c) Microsoft. All rights reserved.

using System;
using System.ComponentModel;
using System.Diagnostics;
using System.IO;
using System.Text.Json;
using System.Threading.Tasks;
using Azure.Identity;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Agents;
using Microsoft.SemanticKernel.Agents.Chat;
using Microsoft.SemanticKernel.Agents.History;
using Microsoft.SemanticKernel.ChatCompletion;
using Microsoft.SemanticKernel.Connectors.AzureOpenAI;

namespace AgentsSample;

public static class Program
{
    public static async Task Main()
    {
        // Load configuration from environment variables or user secrets.
        Settings settings = new();

        Console.WriteLine("Creating kernel...");
        IKernelBuilder builder = Kernel.CreateBuilder();

        builder.AddAzureOpenAIChatCompletion(
            settings.AzureOpenAI.ChatModelDeployment,
            settings.AzureOpenAI.Endpoint,
            new AzureCliCredential());

        Kernel kernel = builder.Build();

        Kernel toolKernel = kernel.Clone();
        toolKernel.Plugins.AddFromType<ClipboardAccess>();


        Console.WriteLine("Defining agents...");

        const string ReviewerName = "Reviewer";
        const string WriterName = "Writer";

        ChatCompletionAgent agentReviewer =
            new()
            {
                Name = ReviewerName,
                Instructions =
                    """
                    Your responsiblity is to review and identify how to improve user provided content.
                    If the user has providing input or direction for content already provided, specify how to address this input.
                    Never directly perform the correction or provide example.
                    Once the content has been updated in a subsequent response, you will review the content again until satisfactory.
                    Always copy satisfactory content to the clipboard using available tools and inform user.

                    RULES:
                    - Only identify suggestions that are specific and actionable.
                    - Verify previous suggestions have been addressed.
                    - Never repeat previous suggestions.
                    """,
                Kernel = toolKernel,
                Arguments = new KernelArguments(new AzureOpenAIPromptExecutionSettings() { FunctionChoiceBehavior = FunctionChoiceBehavior.Auto() })
            };

        ChatCompletionAgent agentWriter =
            new()
            {
                Name = WriterName,
                Instructions =
                    """
                    Your sole responsiblity is to rewrite content according to review suggestions.

                    - Always apply all review direction.
                    - Always revise the content in its entirety without explanation.
                    - Never address the user.
                    """,
                Kernel = kernel,
            };

        KernelFunction selectionFunction =
            AgentGroupChat.CreatePromptFunctionForStrategy(
                $$$"""
                Examine the provided RESPONSE and choose the next participant.
                State only the name of the chosen participant without explanation.
                Never choose the participant named in the RESPONSE.

                Choose only from these participants:
                - {{{ReviewerName}}}
                - {{{WriterName}}}

                Always follow these rules when choosing the next participant:
                - If RESPONSE is user input, it is {{{ReviewerName}}}'s turn.
                - If RESPONSE is by {{{ReviewerName}}}, it is {{{WriterName}}}'s turn.
                - If RESPONSE is by {{{WriterName}}}, it is {{{ReviewerName}}}'s turn.

                RESPONSE:
                {{$lastmessage}}
                """,
                safeParameterNames: "lastmessage");

        const string TerminationToken = "yes";

        KernelFunction terminationFunction =
            AgentGroupChat.CreatePromptFunctionForStrategy(
                $$$"""
                Examine the RESPONSE and determine whether the content has been deemed satisfactory.
                If content is satisfactory, respond with a single word without explanation: {{{TerminationToken}}}.
                If specific suggestions are being provided, it is not satisfactory.
                If no correction is suggested, it is satisfactory.

                RESPONSE:
                {{$lastmessage}}
                """,
                safeParameterNames: "lastmessage");

        ChatHistoryTruncationReducer historyReducer = new(1);

        AgentGroupChat chat =
            new(agentReviewer, agentWriter)
            {
                ExecutionSettings = new AgentGroupChatSettings
                {
                    SelectionStrategy =
                        new KernelFunctionSelectionStrategy(selectionFunction, kernel)
                        {
                            // Always start with the editor agent.
                            InitialAgent = agentReviewer,
                            // Save tokens by only including the final response
                            HistoryReducer = historyReducer,
                            // The prompt variable name for the history argument.
                            HistoryVariableName = "lastmessage",
                            // Returns the entire result value as a string.
                            ResultParser = (result) => result.GetValue<string>() ?? agentReviewer.Name
                        },
                    TerminationStrategy =
                        new KernelFunctionTerminationStrategy(terminationFunction, kernel)
                        {
                            // Only evaluate for editor's response
                            Agents = [agentReviewer],
                            // Save tokens by only including the final response
                            HistoryReducer = historyReducer,
                            // The prompt variable name for the history argument.
                            HistoryVariableName = "lastmessage",
                            // Limit total number of turns
                            MaximumIterations = 12,
                            // Customer result parser to determine if the response is "yes"
                            ResultParser = (result) => result.GetValue<string>()?.Contains(TerminationToken, StringComparison.OrdinalIgnoreCase) ?? false
                        }
                }
            };

        Console.WriteLine("Ready!");

        bool isComplete = false;
        do
        {
            Console.WriteLine();
            Console.Write("> ");
            string input = Console.ReadLine();
            if (string.IsNullOrWhiteSpace(input))
            {
                continue;
            }
            input = input.Trim();
            if (input.Equals("EXIT", StringComparison.OrdinalIgnoreCase))
            {
                isComplete = true;
                break;
            }

            if (input.Equals("RESET", StringComparison.OrdinalIgnoreCase))
            {
                await chat.ResetAsync();
                Console.WriteLine("[Converation has been reset]");
                continue;
            }

            if (input.StartsWith("@", StringComparison.Ordinal) && input.Length > 1)
            {
                string filePath = input.Substring(1);
                try
                {
                    if (!File.Exists(filePath))
                    {
                        Console.WriteLine($"Unable to access file: {filePath}");
                        continue;
                    }
                    input = File.ReadAllText(filePath);
                }
                catch (Exception)
                {
                    Console.WriteLine($"Unable to access file: {filePath}");
                    continue;
                }
            }

            chat.AddChatMessage(new ChatMessageContent(AuthorRole.User, input));

            chat.IsComplete = false;

            try
            {
                await foreach (ChatMessageContent response in chat.InvokeAsync())
                {
                    Console.WriteLine();
                    Console.WriteLine($"{response.AuthorName.ToUpperInvariant()}:{Environment.NewLine}{response.Content}");
                }
            }
            catch (HttpOperationException exception)
            {
                Console.WriteLine(exception.Message);
                if (exception.InnerException != null)
                {
                    Console.WriteLine(exception.InnerException.Message);
                    if (exception.InnerException.Data.Count > 0)
                    {
                        Console.WriteLine(JsonSerializer.Serialize(exception.InnerException.Data, new JsonSerializerOptions() { WriteIndented = true }));
                    }
                }
            }
        } while (!isComplete);
    }

    private sealed class ClipboardAccess
    {
        [KernelFunction]
        [Description("Copies the provided content to the clipboard.")]
        public static void SetClipboard(string content)
        {
            if (string.IsNullOrWhiteSpace(content))
            {
                return;
            }

            using Process clipProcess = Process.Start(
                new ProcessStartInfo
                {
                    FileName = "clip",
                    RedirectStandardInput = true,
                    UseShellExecute = false,
                });

            clipProcess.StandardInput.Write(content);
            clipProcess.StandardInput.Close();
        }
    }
}


# Copyright (c) Microsoft. All rights reserved.

import asyncio
import os

from semantic_kernel.agents import AgentGroupChat, ChatCompletionAgent
from semantic_kernel.agents.strategies.selection.kernel_function_selection_strategy import (
    KernelFunctionSelectionStrategy,
)
from semantic_kernel.agents.strategies.termination.kernel_function_termination_strategy import (
    KernelFunctionTerminationStrategy,
)
from semantic_kernel.connectors.ai.open_ai.services.azure_chat_completion import AzureChatCompletion
from semantic_kernel.contents.chat_message_content import ChatMessageContent
from semantic_kernel.contents.utils.author_role import AuthorRole
from semantic_kernel.functions.kernel_function_decorator import kernel_function
from semantic_kernel.functions.kernel_function_from_prompt import KernelFunctionFromPrompt
from semantic_kernel.kernel import Kernel

###################################################################
# The following sample demonstrates how to create a simple,       #
# agent group chat that utilizes a Reviewer Chat Completion       #
# Agent along with a Writer Chat Completion Agent to              #
# complete a user's task.                                         #
###################################################################


class ClipboardAccess:
    @kernel_function
    def set_clipboard(content: str):
        if not content.strip():
            return

        pyperclip.copy(content)


REVIEWER_NAME = "Reviewer"
COPYWRITER_NAME = "Writer"


def _create_kernel_with_chat_completion(service_id: str) -> Kernel:
    kernel = Kernel()
    kernel.add_service(AzureChatCompletion(service_id=service_id))
    return kernel


async def main():
    agent_reviewer = ChatCompletionAgent(
        service_id=REVIEWER_NAME,
        kernel=_create_kernel_with_chat_completion(REVIEWER_NAME),
        name=REVIEWER_NAME,
        instructions="""
            Your responsiblity is to review and identify how to improve user provided content.
            If the user has providing input or direction for content already provided, specify how to 
            address this input.
            Never directly perform the correction or provide example.
            Once the content has been updated in a subsequent response, you will review the content 
            again until satisfactory.
            Always copy satisfactory content to the clipboard using available tools and inform user.

            RULES:
            - Only identify suggestions that are specific and actionable.
            - Verify previous suggestions have been addressed.
            - Never repeat previous suggestions.
            """,
    )

    agent_writer = ChatCompletionAgent(
        service_id=COPYWRITER_NAME,
        kernel=_create_kernel_with_chat_completion(COPYWRITER_NAME),
        name=COPYWRITER_NAME,
        instructions="""
            Your sole responsiblity is to rewrite content according to review suggestions.

            - Always apply all review direction.
            - Always revise the content in its entirety without explanation.
            - Never address the user.
            """,
    )

    selection_function = KernelFunctionFromPrompt(
        function_name="selection",
        prompt=f"""
        Determine which participant takes the next turn in a conversation based on the the most recent participant.
        State only the name of the participant to take the next turn.
        No participant should take more than one turn in a row.

        Choose only from these participants:
        - {REVIEWER_NAME}
        - {COPYWRITER_NAME}

        Always follow these rules when selecting the next participant:
        - After user input, it is {COPYWRITER_NAME}'s turn.
        - After {COPYWRITER_NAME} replies, it is {REVIEWER_NAME}'s turn.
        - After {REVIEWER_NAME} provides feedback, it is {COPYWRITER_NAME}'s turn.

        History:
        {{{{$history}}}}
        """,
    )

    TERMINATION_KEYWORD = "yes"

    termination_function = KernelFunctionFromPrompt(
        function_name="termination",
        prompt=f"""
            Examine the RESPONSE and determine whether the content has been deemed satisfactory.
            If content is satisfactory, respond with a single word without explanation: {TERMINATION_KEYWORD}.
            If specific suggestions are being provided, it is not satisfactory.
            If no correction is suggested, it is satisfactory.

            RESPONSE:
            {{{{$history}}}}
            """,
    )

    chat = AgentGroupChat(
        agents=[agent_writer, agent_reviewer],
        selection_strategy=KernelFunctionSelectionStrategy(
            function=selection_function,
            kernel=_create_kernel_with_chat_completion("selection"),
            result_parser=lambda result: str(result.value[0]) if result.value is not None else COPYWRITER_NAME,
            agent_variable_name="agents",
            history_variable_name="history",
        ),
        termination_strategy=KernelFunctionTerminationStrategy(
            agents=[agent_reviewer],
            function=termination_function,
            kernel=_create_kernel_with_chat_completion("termination"),
            result_parser=lambda result: TERMINATION_KEYWORD in str(result.value[0]).lower(),
            history_variable_name="history",
            maximum_iterations=10,
        ),
    )

    is_complete: bool = False
    while not is_complete:
        user_input = input("User:> ")
        if not user_input:
            continue

        if user_input.lower() == "exit":
            is_complete = True
            break

        if user_input.lower() == "reset":
            await chat.reset()
            print("[Conversation has been reset]")
            continue

        if user_input.startswith("@") and len(input) > 1:
            file_path = input[1:]
            try:
                if not os.path.exists(file_path):
                    print(f"Unable to access file: {file_path}")
                    continue
                with open(file_path) as file:
                    user_input = file.read()
            except Exception:
                print(f"Unable to access file: {file_path}")
                continue

        await chat.add_chat_message(ChatMessageContent(role=AuthorRole.USER, content=user_input))

        async for response in chat.invoke():
            print(f"# {response.role} - {response.name or '*'}: '{response.content}'")

        if chat.is_complete:
            is_complete = True
            break


if __name__ == "__main__":
    asyncio.run(main())

Agents are currently unavailable in Java.