autogen/dotnet/samples/AgentChat/AutoGen.Basic.Sample/Example05_Dalle_And_GPT4V.cs

// Copyright (c) Microsoft Corporation. All rights reserved.
// Example05_Dalle_And_GPT4V.cs

using AutoGen.Core;
using AutoGen.OpenAI;
using AutoGen.OpenAI.Extension;
using FluentAssertions;
using OpenAI;
using OpenAI.Images;

public partial class Example05_Dalle_And_GPT4V
{
    private readonly OpenAIClient openAIClient;

    public Example05_Dalle_And_GPT4V(OpenAIClient openAIClient)
    {
        this.openAIClient = openAIClient;
    }

    /// <summary>
    /// Generate image from prompt using DALL-E.
    /// </summary>
    /// <param name="prompt">prompt with feedback</param>
    /// <returns></returns>
    [Function]
    public async Task<string> GenerateImage(string prompt)
    {
        // TODO
        // generate image from prompt using DALL-E
        // and return url.
        var option = new ImageGenerationOptions
        {
            Size = GeneratedImageSize.W1024xH1024,
            Style = GeneratedImageStyle.Vivid,
        };

        var imageResponse = await openAIClient.GetImageClient("dall-e-3").GenerateImageAsync(prompt, option);
        var imageUrl = imageResponse.Value.ImageUri.OriginalString;

        return $@"// ignore this line [IMAGE_GENERATION]
The image is generated from prompt {prompt}

{imageUrl}";
    }

    public static async Task RunAsync()
    {
        // This example shows how to use DALL-E and GPT-4V to generate image from prompt and feedback.
        // The DALL-E agent will generate image from prompt.
        // The GPT-4V agent will provide feedback to DALL-E agent to help it generate better image.
        // The conversation will be terminated when the image satisfies the condition.
        // The image will be saved to image.jpg in current directory.

        // get OpenAI Key and create config
        var openAIKey = Environment.GetEnvironmentVariable("OPENAI_API_KEY") ?? throw new Exception("Please set OPENAI_API_KEY environment variable.");
        var openAIClient = new OpenAIClient(openAIKey);
        var instance = new Example05_Dalle_And_GPT4V(openAIClient);
        var imagePath = Path.Combine("resource", "images", "background.png");
        if (File.Exists(imagePath))
        {
            File.Delete(imagePath);
        }

        var generateImageFunctionMiddleware = new FunctionCallMiddleware(
            functions: [instance.GenerateImageFunctionContract],
            functionMap: new Dictionary<string, Func<string, Task<string>>>
            {
                { nameof(GenerateImage), instance.GenerateImageWrapper },
            });
        var dalleAgent = new OpenAIChatAgent(
            chatClient: openAIClient.GetChatClient("gpt-4o-mini"),
            name: "dalle",
            systemMessage: "You are a DALL-E agent that generate image from prompt, when conversation is terminated, return the most recent image url")
            .RegisterMessageConnector()
            .RegisterStreamingMiddleware(generateImageFunctionMiddleware)
            .RegisterMiddleware(async (msgs, option, agent, ct) =>
            {
                if (msgs.Any(msg => msg.GetContent()?.ToLower().Contains("approve") is true))
                {
                    return new TextMessage(Role.Assistant, $"The image satisfies the condition, conversation is terminated. {GroupChatExtension.TERMINATE}");
                }

                var msgsWithoutImage = msgs.Where(msg => msg is not ImageMessage).ToList();
                var reply = await agent.GenerateReplyAsync(msgsWithoutImage, option, ct);

                if (reply.GetContent() is string content && content.Contains("IMAGE_GENERATION"))
                {
                    var imageUrl = content.Split("\n").Last();
                    var imageMessage = new ImageMessage(Role.Assistant, imageUrl, from: reply.From, mimeType: "image/png");

                    Console.WriteLine($"download image from {imageUrl} to {imagePath}");
                    var httpClient = new HttpClient();
                    var imageBytes = await httpClient.GetByteArrayAsync(imageUrl, ct);
                    File.WriteAllBytes(imagePath, imageBytes);

                    return imageMessage;
                }
                else
                {
                    return reply;
                }
            })
            .RegisterPrintMessage();

        var gpt4VAgent = new OpenAIChatAgent(
            chatClient: openAIClient.GetChatClient("gpt-4o-mini"),
            name: "gpt-4o-mini",
            systemMessage: @"You are a critism that provide feedback to DALL-E agent.
Carefully check the image generated by DALL-E agent and provide feedback.
If the image satisfies the condition, then say [APPROVE].
Otherwise, provide detailed feedback to DALL-E agent so it can generate better image.

The image should satisfy the following conditions:
- There should be a cat and a mouse in the image
- The cat should be chasing after the mouse")
            .RegisterMessageConnector()
            .RegisterPrintMessage();

        await gpt4VAgent.InitiateChatAsync(
            receiver: dalleAgent,
            message: "Hey dalle, please generate image from prompt: English short hair blue cat chase after a mouse",
            maxRound: 10);

        File.Exists(imagePath).Should().BeTrue();
    }
}