chore: initial commit

src/AltGen.API/AltGen.API.csproj

@@ -0,0 +1,20 @@
+<Project Sdk="Microsoft.NET.Sdk.Web">
+
+  <PropertyGroup>
+    <TargetFramework>net9.0</TargetFramework>
+    <Nullable>enable</Nullable>
+    <ImplicitUsings>enable</ImplicitUsings>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="Microsoft.AspNetCore.OpenApi" Version="9.0.1" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <InternalsVisibleTo Include="$(AssemblyName).Tests" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <EmbeddedResource Include="Resources\**\*" />
+  </ItemGroup>
+</Project>

src/AltGen.API/AltGen.API.http

@@ -0,0 +1,6 @@
+@AltGen.API_HostAddress = http://localhost:5005
+
+GET {{AltGen.API_HostAddress}}/weatherforecast/
+Accept: application/json
+
+###

src/AltGen.API/Common/AltGenException.cs

@@ -0,0 +1,5 @@
+namespace AltGen.API.Common;
+
+class AltGenException(string message) : Exception(message)
+{
+}

src/AltGen.API/Common/ValidationResultsExtensions.cs

@@ -0,0 +1,14 @@
+namespace AltGen.API.Common;
+
+static class ValidationResultsExtensions
+{
+  public static Dictionary<string, string[]> ToErrors(this IEnumerable<ValidationResult> results)
+  {
+    // group by the first member name and select the error message
+    return results.GroupBy(static r => r.MemberNames.First())
+      .ToDictionary(
+        static r => r.Key,
+        static r => r.Select(static e => e.ErrorMessage!).ToArray()
+      );
+  }
+}

src/AltGen.API/Generate/GenerateEndpoint.cs

@@ -0,0 +1,27 @@
+namespace AltGen.API.Generate;
+
+static class GenerateEndpoint
+{
+  public static async Task<IResult> HandleAsync([AsParameters] GenerateRequest request, [FromServices] IAltTextProviderFactory factory)
+  {
+    var validationResults = request.Validate(new ValidationContext(request));
+
+    if (validationResults.Any())
+    {
+      return Results.ValidationProblem(validationResults.ToErrors());
+    }
+
+    var provider = factory.Create(request.Provider);
+
+    var imageStream = new MemoryStream();
+    await request.File.CopyToAsync(imageStream);
+
+    var altText = await provider.GenerateAltTextAsync(
+      request.ProviderKey,
+      request.File.ContentType,
+      imageStream
+    );
+
+    return Results.Ok(new GenerateResponse(altText));
+  }
+}

src/AltGen.API/Generate/GenerateRequest.cs

@@ -0,0 +1,47 @@
+namespace AltGen.API.Generate;
+
+record GenerateRequest(
+  [FromForm]
+  string Provider,
+  [FromForm]
+  string ProviderKey,
+  IFormFile File
+) : IValidatableObject
+{
+  public IEnumerable<ValidationResult> Validate(ValidationContext validationContext)
+  {
+    if (string.IsNullOrWhiteSpace(Provider))
+    {
+      yield return new ValidationResult($"The {nameof(Provider)} field is required.", [nameof(Provider)]);
+    }
+
+    if (string.IsNullOrWhiteSpace(Provider) is false && LLMProvider.IsValid(Provider) is false)
+    {
+      yield return new ValidationResult($"The {nameof(Provider)} field is invalid.", [nameof(Provider)]);
+    }
+
+    if (string.IsNullOrWhiteSpace(ProviderKey))
+    {
+      yield return new ValidationResult($"The {nameof(ProviderKey)} field is required.", [nameof(ProviderKey)]);
+    }
+
+    // TODO: Probably need to consider upper bound for file size
+    // this also needs to take into account provider-specific limits
+    if (File.Length is 0)
+    {
+      yield return new ValidationResult($"The {nameof(File)} has no content.", [nameof(File)]);
+    }
+
+    if (File.Length > 0)
+    {
+      var extension = Path.GetExtension(File.FileName);
+
+      // TODO: prob needs to be specific to the provider
+      // different LLMs support different image formats
+      if (extension is not ".jpeg" and not ".jpg" and not ".png")
+      {
+        yield return new ValidationResult($"The {nameof(File)} must be a valid image file.", [nameof(File)]);
+      }
+    }
+  }
+}

src/AltGen.API/Generate/GenerateResponse.cs

@@ -0,0 +1,3 @@
+namespace AltGen.API.Generate;
+
+record GenerateResponse(string AltText);

src/AltGen.API/Generate/Providers/AltTextProviderFactory.cs

@@ -0,0 +1,15 @@
+namespace AltGen.API.Generate.Providers;
+
+class AltTextProviderFactory(IServiceProvider serviceProvider) : IAltTextProviderFactory
+{
+  readonly IServiceProvider _serviceProvider = serviceProvider;
+
+  public IAltTextProvider Create(string provider)
+  {
+    return provider switch
+    {
+      LLMProvider.Gemini => _serviceProvider.GetRequiredKeyedService<IAltTextProvider>(LLMProvider.Gemini),
+      _ => throw new NotSupportedException($"The provider '{provider}' is not supported.")
+    };
+  }
+}

src/AltGen.API/Generate/Providers/Gemini/Candidate.cs

@@ -0,0 +1,3 @@
+namespace AltGen.API.Generate.Providers.Gemini;
+
+record Candidate(Content Content);

src/AltGen.API/Generate/Providers/Gemini/Content.cs

@@ -0,0 +1,3 @@
+namespace AltGen.API.Generate.Providers.Gemini;
+
+record Content(Part[] Parts, string Role);

src/AltGen.API/Generate/Providers/Gemini/GeminiAltTextProvider.cs

@@ -0,0 +1,11 @@
+namespace AltGen.API.Generate.Providers.Gemini;
+
+class GeminiAltTextProvider(IGeminiService geminiService) : IAltTextProvider
+{
+  readonly IGeminiService _geminiService = geminiService;
+
+  public Task<string> GenerateAltTextAsync(string providerKey, string mimeType, MemoryStream image)
+  {
+    return _geminiService.GenerateContentAsync(providerKey, mimeType, image);
+  }
+}

src/AltGen.API/Generate/Providers/Gemini/GeminiRequest.cs

@@ -0,0 +1,6 @@
+namespace AltGen.API.Generate.Providers.Gemini;
+
+record GeminiRequest(
+  Content SystemInstruction,
+  Content[] Contents
+);

src/AltGen.API/Generate/Providers/Gemini/GeminiResponse.cs

@@ -0,0 +1,3 @@
+namespace AltGen.API.Generate.Providers.Gemini;
+
+record GeminiResponse(Candidate[] Candidates);

src/AltGen.API/Generate/Providers/Gemini/GeminiRole.cs

@@ -0,0 +1,8 @@
+namespace AltGen.API.Generate.Providers.Gemini;
+
+static class GeminiRole
+{
+  public const string System = "system";
+  public const string User = "user";
+  public const string Model = "model";
+}

src/AltGen.API/Generate/Providers/Gemini/GeminiService.cs

@@ -0,0 +1,79 @@
+namespace AltGen.API.Generate.Providers.Gemini;
+
+class GeminiService(HttpClient httpClient) : IGeminiService
+{
+  const string BaseUri = "https://generativelanguage.googleapis.com/v1beta/models";
+  const string Method = "generateContent";
+  const string ModelId = "gemini-1.5-flash";
+  const string ApiKeyQueryKey = "key";
+
+  // TODO: Use Lazy<T> to load the prompt resource
+  string _prompt = "";
+  string Prompt
+  {
+    get
+    {
+      if (string.IsNullOrWhiteSpace(_prompt))
+      {
+        var resource = Assembly.GetExecutingAssembly()
+          .GetManifestResourceStream("AltGen.API.Resources.prompt.txt") ?? throw new AltGenException("Failed to load the prompt resource.");
+        using var reader = new StreamReader(resource);
+        _prompt = reader.ReadToEnd();
+      }
+
+      return _prompt;
+    }
+  }
+
+  static readonly JsonSerializerOptions Options = new()
+  {
+    PropertyNameCaseInsensitive = true,
+    PropertyNamingPolicy = JsonNamingPolicy.CamelCase,
+    Converters = { new PartConverter() }
+  };
+  readonly HttpClient _httpClient = httpClient;
+
+  public async Task<string> GenerateContentAsync(string providerKey, string mimeType, MemoryStream image)
+  {
+    var requestUri = GetRequestUri(providerKey);
+    var imageBase64 = ConvertToBase64(image);
+
+    var request = new GeminiRequest(
+      new Content([new TextPart(Prompt)], GeminiRole.System),
+      [new Content(
+        [
+          new TextPart(""),
+          new InlineDataPart(new InlineData(mimeType, imageBase64))
+        ],
+        GeminiRole.User
+      )]
+    );
+
+    var response = await _httpClient.PostAsJsonAsync(requestUri, request, Options);
+    var content = await response.Content.ReadAsStringAsync();
+
+    if (response.IsSuccessStatusCode is false)
+    {
+      throw new AltGenException("Failed to generate alt text.");
+    }
+
+    var geminiResponse = JsonSerializer.Deserialize<GeminiResponse>(content, Options) ?? throw new AltGenException("Failed to deserialize the Gemini response.");
+    var firstCandidate = geminiResponse.Candidates.FirstOrDefault() ?? throw new AltGenException("No candidates found in the Gemini response.");
+    var altText = firstCandidate.Content.Parts
+      .OfType<TextPart>()
+      .Aggregate(new StringBuilder(), static (sb, part) => sb.Append(part.Text))
+      .ToString();
+
+    return altText;
+  }
+
+  static string GetRequestUri(string providerKey)
+  {
+    return $"{BaseUri}/{ModelId}:{Method}/?{ApiKeyQueryKey}={providerKey}";
+  }
+
+  static string ConvertToBase64(MemoryStream image)
+  {
+    return Convert.ToBase64String(image.ToArray());
+  }
+}

src/AltGen.API/Generate/Providers/Gemini/IGeminiService.cs

@@ -0,0 +1,6 @@
+namespace AltGen.API.Generate.Providers.Gemini;
+
+interface IGeminiService
+{
+  Task<string> GenerateContentAsync(string providerKey, string mimeType, MemoryStream image);
+}

src/AltGen.API/Generate/Providers/Gemini/Part.cs

@@ -0,0 +1,9 @@
+namespace AltGen.API.Generate.Providers.Gemini;
+
+record Part();
+
+record TextPart(string Text) : Part;
+
+record InlineDataPart(InlineData InlineData) : Part;
+
+record InlineData(string MimeType, string Data);

src/AltGen.API/Generate/Providers/Gemini/PartConverter.cs

@@ -0,0 +1,27 @@
+namespace AltGen.API.Generate.Providers.Gemini;
+
+class PartConverter : JsonConverter<Part>
+{
+  public override Part? Read(ref Utf8JsonReader reader, Type typeToConvert, JsonSerializerOptions options)
+  {
+    using var doc = JsonDocument.ParseValue(ref reader);
+    var root = doc.RootElement;
+
+    if (root.TryGetProperty("text", out var text))
+    {
+      return JsonSerializer.Deserialize<TextPart>(root.GetRawText(), options);
+    }
+
+    if (root.TryGetProperty("inlineData", out var inlineData))
+    {
+      return JsonSerializer.Deserialize<InlineDataPart>(root.GetRawText(), options);
+    }
+
+    throw new JsonException("Invalid part type.");
+  }
+
+  public override void Write(Utf8JsonWriter writer, Part value, JsonSerializerOptions options)
+  {
+    JsonSerializer.Serialize(writer, value, value.GetType(), options);
+  }
+}

src/AltGen.API/Generate/Providers/IAltTextProvider.cs

@@ -0,0 +1,6 @@
+namespace AltGen.API.Generate.Providers;
+
+interface IAltTextProvider
+{
+  Task<string> GenerateAltTextAsync(string providerKey, string mimeType, MemoryStream image);
+}

src/AltGen.API/Generate/Providers/IAltTextProviderFactory.cs

@@ -0,0 +1,6 @@
+namespace AltGen.API.Generate.Providers;
+
+interface IAltTextProviderFactory
+{
+  IAltTextProvider Create(string provider);
+}

src/AltGen.API/Generate/Providers/LLMProvider.cs

@@ -0,0 +1,11 @@
+namespace AltGen.API.Generate.Providers;
+
+static class LLMProvider
+{
+  public const string Gemini = "Gemini";
+
+  public static bool IsValid(string provider)
+  {
+    return provider is Gemini;
+  }
+}

src/AltGen.API/Program.cs

@@ -0,0 +1,26 @@
+var builder = WebApplication.CreateBuilder(args);
+
+builder.Services.AddProblemDetails();
+builder.Services.AddOpenApi();
+
+builder.Services.AddHttpClient<IGeminiService, GeminiService>();
+builder.Services.AddSingleton<IAltTextProviderFactory, AltTextProviderFactory>();
+builder.Services.AddKeyedSingleton<IAltTextProvider, GeminiAltTextProvider>(LLMProvider.Gemini);
+
+var app = builder.Build();
+
+if (app.Environment.IsDevelopment())
+{
+  app.MapOpenApi();
+}
+
+app.UseHttpsRedirection();
+
+app.UseStatusCodePages();
+
+app.MapPost("/generate", GenerateEndpoint.HandleAsync).DisableAntiforgery();
+
+app.Run();
+
+[ExcludeFromCodeCoverage]
+public partial class Program { }

src/AltGen.API/Properties/launchSettings.json

@@ -0,0 +1,23 @@
+{
+  "$schema": "https://json.schemastore.org/launchsettings.json",
+  "profiles": {
+    "https": {
+      "commandName": "Project",
+      "dotnetRunMessages": true,
+      "launchBrowser": false,
+      "applicationUrl": "https://localhost:7297;http://localhost:5005",
+      "environmentVariables": {
+        "ASPNETCORE_ENVIRONMENT": "Development"
+      }
+    },
+    "http": {
+      "commandName": "Project",
+      "dotnetRunMessages": true,
+      "launchBrowser": false,
+      "applicationUrl": "http://localhost:5005",
+      "environmentVariables": {
+        "ASPNETCORE_ENVIRONMENT": "Development"
+      }
+    },
+  }
+}

src/AltGen.API/Resources/prompt.txt

@@ -0,0 +1,17 @@
+Generate descriptive alt text that captures the essential visual information of the image. Follow these guidelines:
+
+1. Be concise, but comprehensive (1-2 sentences)
+2. Describe the most important visual elements
+3. Convey the image's purpose or key message
+4. Use objective language
+5. Avoid redundant phrases
+
+Priority details to include:
+- Main subject(s)
+- Action or context
+- Color or distinctive visual characteristics
+- Emotional tone or artistic intent
+
+Exclude unnecessary details like background minutiae or decorative elements unless they are crucial to understanding the image.
+
+Please DO NOT respond with anything over than the alt text.

src/AltGen.API/Usings.cs

@@ -0,0 +1,13 @@
+global using System.ComponentModel.DataAnnotations;
+global using System.Diagnostics.CodeAnalysis;
+global using System.Reflection;
+global using System.Text;
+global using System.Text.Json;
+global using System.Text.Json.Serialization;
+
+global using AltGen.API.Common;
+global using AltGen.API.Generate;
+global using AltGen.API.Generate.Providers;
+global using AltGen.API.Generate.Providers.Gemini;
+
+global using Microsoft.AspNetCore.Mvc;