Merge pull request #4 from aghimir3/issues/5657-itexttoimage-ag

.NET Add Style and Quality Parameters Support with Execution Settings

dotnet/src/Connectors/Connectors.OpenAI/TextToImage/OpenAITextToImageService.cs

@@ -3,6 +3,7 @@
 using System;
 using System.Collections.Generic;
 using System.Diagnostics.CodeAnalysis;
+using System.Linq;
 using System.Net.Http;
 using System.Text.Json;
 using System.Threading;
@@ -94,9 +95,72 @@ public Task<string> GenerateImageAsync(string description, int width, int height
     }
 
     /// <inheritdoc/>
-    public Task<IReadOnlyList<ImageContent>> GetImageContentsAsync(TextContent input, PromptExecutionSettings? executionSettings = null, Kernel? kernel = null, CancellationToken cancellationToken = default)
+    public async Task<IReadOnlyList<ImageContent>> GetImageContentsAsync(
+        TextContent input,
+        PromptExecutionSettings? executionSettings = null,
+        Kernel? kernel = null,
+        CancellationToken cancellationToken = default)
     {
-        throw new NotImplementedException();
+        // Ensure the input is valid
+        Verify.NotNull(input);
+
+        // Convert the generic execution settings to OpenAI-specific settings
+        var imageSettings = OpenAITextToImageExecutionSettings.FromExecutionSettings(executionSettings);
+
+        // Determine the size of the image based on the width and height settings
+        var size = (imageSettings.Width, imageSettings.Height) switch
+        {
+            (256, 256) => "256x256",
+            (512, 512) => "512x512",
+            (1024, 1024) => "1024x1024",
+            (1792, 1024) => "1792x1024",
+            (1024, 1792) => "1024x1792",
+            _ => throw new NotSupportedException($"The provided size is not supported: {imageSettings.Width}x{imageSettings.Height}")
+        };
+
+        // Validate quality and style
+        var supportedQualities = new[] { "standard", "hd" };
+        var supportedStyles = new[] { "vivid", "natural" };
+
+        if (!string.IsNullOrEmpty(imageSettings.Quality) && !supportedQualities.Contains(imageSettings.Quality))
+        {
+            throw new NotSupportedException($"The provided quality '{imageSettings.Quality}' is not supported.");
+        }
+
+        if (!string.IsNullOrEmpty(imageSettings.Style) && !supportedStyles.Contains(imageSettings.Style))
+        {
+            throw new NotSupportedException($"The provided style '{imageSettings.Style}' is not supported.");
+        }
+
+        // Create the request body for the image generation
+        var requestBody = JsonSerializer.Serialize(new TextToImageRequest
+        {
+            Model = imageSettings.ModelId ?? this._modelId,
+            Prompt = input.Text ?? string.Empty,
+            Size = size,
+            Count = imageSettings.ImageCount ?? 1,
+            Quality = imageSettings.Quality ?? "standard",
+            Style = imageSettings.Style ?? "vivid"
+        });
+
+        // Execute the request using the core client and return Image objects
+        var imageStrings = await this._core.ExecuteImageGenerationRequestAsync(OpenAIEndpoint, requestBody, x => x.Url ?? x.AsBase64, cancellationToken).ConfigureAwait(false);
+
+        // Convert the strings to ImageContent objects
+        var images = new List<ImageContent>();
+        foreach (var imageString in imageStrings)
+        {
+            if (Uri.TryCreate(imageString, UriKind.Absolute, out var uriResult) && (uriResult.Scheme == Uri.UriSchemeHttp || uriResult.Scheme == Uri.UriSchemeHttps))
+            {
+                images.Add(new ImageContent(uriResult));
+            }
+            else
+            {
+                images.Add(new ImageContent($"data:;base64,{imageString}"));
+            }
+        }
+
+        return images.AsReadOnly();
     }
 
     private async Task<string> GenerateImageAsync(

dotnet/src/Connectors/Connectors.OpenAI/TextToImage/TextToImageRequest.cs

@@ -39,4 +39,16 @@ internal sealed class TextToImageRequest
     /// </summary>
     [JsonPropertyName("response_format")]
     public string Format { get; set; } = "url";
+
+    /// <summary>
+    /// Image quality, "standard" or "hd"
+    /// </summary>
+    [JsonPropertyName("quality")]
+    public string Quality { get; set; } = "standard";
+
+    /// <summary>
+    /// Image style, "vivid" or "natural"
+    /// </summary>
+    [JsonPropertyName("style")]
+    public string Style { get; set; } = "vivid";
 }

dotnet/src/Connectors/Connectors.UnitTests/OpenAI/TextToImage/OpenAITextToImageServiceTests.cs

@@ -1,10 +1,12 @@
 // Copyright (c) Microsoft. All rights reserved.
 
 using System;
+using System.Linq;
 using System.Net.Http;
 using System.Text;
 using System.Threading.Tasks;
 using Microsoft.Extensions.Logging;
+using Microsoft.SemanticKernel;
 using Microsoft.SemanticKernel.Connectors.OpenAI;
 using Moq;
 using Xunit;
@@ -81,6 +83,122 @@ public async Task GenerateImageWorksCorrectlyAsync(int width, int height, bool e
         }
     }
 
+    [Fact]
+    public async Task GetImageContentsAsyncWithValidInputReturnsImageContentsAsync()
+    {
+        // Arrange
+        var service = new OpenAITextToImageService("api-key", "organization", "dall-e-3", this._httpClient);
+        Assert.Equal("dall-e-3", service.Attributes["ModelId"]);
+
+        this._messageHandlerStub.ResponseToReturn = new HttpResponseMessage(System.Net.HttpStatusCode.OK)
+        {
+            Content = new StringContent("""
+                                    {
+                                        "created": 1702575371,
+                                        "data": [
+                                            {
+                                                "url": "https://image-url"
+                                            }
+                                        ]
+                                    }
+                                    """, Encoding.UTF8, "application/json")
+        };
+
+        var input = new TextContent("A cute baby sea otter");
+        var executionSettings = new OpenAITextToImageExecutionSettings
+        {
+            Width = 1024,
+            Height = 1024,
+            Quality = "hd",
+            Style = "natural",
+            ImageCount = 1
+        };
+
+        // Act
+        var result = await service.GetImageContentsAsync(input, executionSettings);
+
+        // Assert
+        Assert.NotNull(result);
+        Assert.Single(result);
+        Assert.Equal(new Uri("https://image-url"), result[0].Uri);
+    }
+
+    [Theory]
+    [InlineData(1024, 1024, "hd", "natural", 1, "https://image-url", false)]
+    [InlineData(123, 456, "hd", "natural", 1, "", true)]
+    [InlineData(1024, 1024, "hd", "natural", 2, "https://image-url1|https://image-url2", false)]
+    [InlineData(1024, 1024, "ultra", "natural", 1, "", true)]
+    [InlineData(1024, 1024, "hd", "artistic", 1, "", true)]
+    public async Task GetImageContentsReturnsExpectedResultsAsync(
+        int width,
+        int height,
+        string quality,
+        string style,
+        int imageCount,
+        string expectedUrls,
+        bool expectException)
+    {
+        // Arrange
+        var service = new OpenAITextToImageService("api-key", "organization", "dall-e-3", this._httpClient);
+
+        if (!expectException)
+        {
+            var urls = expectedUrls.Split('|').Select(url =>
+            {
+                return url.StartsWith("http", StringComparison.OrdinalIgnoreCase) ?
+                    $"{{ \"url\": \"{url}\" }}" :
+                    $"{{ \"b64_json\": \"{url}\" }}";
+            });
+            var jsonResponse = $"{{ \"created\": 1702575371, \"data\": [ {string.Join(",", urls)} ] }}";
+
+            this._messageHandlerStub.ResponseToReturn = new HttpResponseMessage(System.Net.HttpStatusCode.OK)
+            {
+                Content = new StringContent(jsonResponse, Encoding.UTF8, "application/json")
+            };
+        }
+
+        var input = new TextContent("A picturesque landscape");
+        var executionSettings = new OpenAITextToImageExecutionSettings
+        {
+            Width = width,
+            Height = height,
+            Quality = quality,
+            Style = style,
+            ImageCount = imageCount
+        };
+
+        // Act & Assert
+        if (expectException)
+        {
+            await Assert.ThrowsAsync<NotSupportedException>(async () =>
+            {
+                await service.GetImageContentsAsync(input, executionSettings);
+            });
+        }
+        else
+        {
+            var result = await service.GetImageContentsAsync(input, executionSettings);
+
+            Assert.NotNull(result);
+            Assert.Equal(imageCount, result.Count);
+
+            var expectedUrlList = expectedUrls.Split('|').ToList();
+            for (int i = 0; i < result.Count; i++)
+            {
+                if (Uri.TryCreate(expectedUrlList[i], UriKind.Absolute, out var uriResult) &&
+                    (uriResult.Scheme == Uri.UriSchemeHttp || uriResult.Scheme == Uri.UriSchemeHttps))
+                {
+                    Assert.Equal(uriResult, result[i].Uri);
+                }
+                else
+                {
+                    Assert.StartsWith("data:;base64,", result[i].DataUri);
+                    Assert.Contains(expectedUrlList[i], result[i].DataUri);
+                }
+            }
+        }
+    }
+
     public void Dispose()
     {
         this._httpClient.Dispose();