Firstly, following the installation guide to install AutoGen packages.

Then you can start with the following code snippet to create a conversable agent and chat with it.

using AutoGen;
using AutoGen.OpenAI;

var openAIKey = Environment.GetEnvironmentVariable("OPENAI_API_KEY") ?? throw new Exception("Please set OPENAI_API_KEY environment variable.");
var gpt35Config = new OpenAIConfig(openAIKey, "gpt-3.5-turbo");

var assistantAgent = new AssistantAgent(
    name: "assistant",
    systemMessage: "You are an assistant that help user to do some tasks.",
    llmConfig: new ConversableAgentConfig
    {
        Temperature = 0,
        ConfigList = [gpt35Config],
    })
    .RegisterPrintMessage(); // register a hook to print message nicely to console

// set human input mode to ALWAYS so that user always provide input
var userProxyAgent = new UserProxyAgent(
    name: "user",
    humanInputMode: ConversableAgent.HumanInputMode.ALWAYS)
    .RegisterPrintMessage();

// start the conversation
await userProxyAgent.InitiateChatAsync(
    receiver: assistantAgent,
    message: "Hey assistant, please do me a favor.",
    maxRound: 10);

You can find more examples under the sample project.

• ConversableAgent

  • function call
  • code execution (dotnet only, powered by dotnet-interactive)

• Agent communication

  • Two-agent chat
  • Group chat

• Enhanced LLM Inferences

• Exclusive for dotnet

  • Source generator for type-safe function definition generation

• Add link to Discord channel in nuget's readme.md
• Document improvements

• Rename Workflow to Graph
• Rename AddInitializeMessage to SendIntroduction
• Rename SequentialGroupChat to RoundRobinGroupChat

• Refactor over @AutoGen.Message and introducing TextMessage, ImageMessage, MultiModalMessage and so on. PR #1676
• Add AutoGen.SemanticKernel to support seamless integration with Semantic Kernel
• Move the agent contract abstraction to AutoGen.Core package. The AutoGen.Core package provides the abstraction for message type, agent and group chat and doesn't contain dependencies over Azure.AI.OpenAI or Semantic Kernel. This is useful when you want to leverage AutoGen's abstraction only and want to avoid introducing any other dependencies.
• Move GPTAgent, OpenAIChatAgent and all openai-dependencies to AutoGen.OpenAI

• Fix #1804
• Streaming support for IAgent #1656
• Streaming support for middleware via MiddlewareStreamingAgent #1656
• Graph chat support with conditional transition workflow #1761
• AutoGen.SourceGenerator: Generate FunctionContract from FunctionAttribute #1736

• Add AutoGen.LMStudio to support comsume openai-like API from LMStudio local server

• Add MiddlewareAgent
• Use MiddlewareAgent to implement existing agent hooks (RegisterPreProcess, RegisterPostProcess, RegisterReply)
• Remove AutoReplyAgent, PreProcessAgent, PostProcessAgent because they are replaced by MiddlewareAgent

• Simplify IAgent interface by removing ChatLLM Property
• Add GenerateReplyOptions to IAgent.GenerateReplyAsync which allows user to specify or override the options when generating reply

• Move out dependency of Semantic Kernel
• Add type IChatLLM as connector to LLM

• In AutoGen.SourceGenerator, rename FunctionAttribution to FunctionAttribute
• In AutoGen, refactor over ConversationAgent, UserProxyAgent, and AssistantAgent

• update Azure.OpenAI.AI to 1.0.0-beta.12
• update Semantic kernel to 1.0.1