chore: initial commit

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);
+  }
+}