Fluent API Pattern

The Fluent API pattern in AI Dev Kit follows a consistent, beginner-friendly design that makes AI integration feel natural.

The Pattern

All generative AI requests follow this structure:

[Input].GENXxx().SetYyy(...).ExecuteAsync()

1. Input (Entry Point)

The input can be various types:

// String
"Hello, AI!".GENCompletion()

// Texture2D
texture.GENInpaint("Add clouds")

// AudioClip
audioClip.GENTranscript()

// Prompt object
prompt.GENResponse()

// Message object
message.GENCompletion()

// ConversationItem
conversationItem.GENResponse()

2. Request Factory (.GENXxx())

The .GENXxx() methods create strongly-typed request objects:

ChatCompletionRequest request = "Hello".GENCompletion();
ImageGenerationRequest request = "Cat".GENImage();
SpeechGenerationRequest request = "Hi".GENSpeech();

Important: These methods do NOT perform any I/O or network calls. They only create configuration objects.

3. Configuration (.SetYyy())

Chain configuration methods to customize the request:

var request = "Explain AI"
    .GENCompletion()
    .SetModel(OpenAIModel.GPT4o)           // Model selection
    .SetTemperature(0.7f)                  // Creativity
    .SetMaxTokens(500)                     // Response length
    .SetSystemMessage("You are helpful"); // Context

4. Execution (.ExecuteAsync())

This is when the actual API call happens:

// Network request is sent NOW
string result = await request.ExecuteAsync();

Why This Pattern?

✅ Readable

// Clear, self-documenting code
var response = await "Explain photosynthesis"
    .GENCompletion()
    .SetModel(OpenAIModel.GPT4o)
    .ExecuteAsync();

✅ Discoverable

Type a dot after any string, texture, or audio clip and IntelliSense shows you all available .GENXxx() methods.

✅ Type-Safe

// Compile-time type checking
Texture2D image = await "Cat".GENImage().ExecuteAsync();
AudioClip audio = await "Hello".GENSpeech().ExecuteAsync();
string text = await "Hi".GENCompletion().ExecuteAsync();

✅ Flexible

You can store and reuse request objects:

// Create request
var request = "Explain {topic}"
    .GENCompletion()
    .SetModel(OpenAIModel.GPT4o);

// Execute multiple times with different inputs
string ai = await request.SetInput("AI").ExecuteAsync();
string ml = await request.SetInput("ML").ExecuteAsync();

Common Patterns

Pattern 1: One-liner

For quick, simple requests:

string response = await "Hello".GENCompletion().ExecuteAsync();

Pattern 2: Stored Configuration

When you need to reuse settings:

var imageRequest = "placeholder"
    .GENImage()
    .SetModel(OpenAIModel.DallE3)
    .SetSize(ImageSize._1024x1024)
    .SetQuality(ImageQuality.HD);

Texture2D logo = await imageRequest.SetPrompt("Company logo").ExecuteAsync();
Texture2D banner = await imageRequest.SetPrompt("Hero banner").ExecuteAsync();

Pattern 3: Async/Await

Always use await or .Forget():

// ✅ Correct - await the result
string text = await "Hello".GENCompletion().ExecuteAsync();

// ✅ Correct - fire and forget
"Hello".GENCompletion().ExecuteAsync().Forget();

// ❌ Wrong - compiler warning
"Hello".GENCompletion().ExecuteAsync();

Pattern 4: Error Handling

Wrap in try-catch for robust error handling:

try
{
    var result = await "Hello".GENCompletion().ExecuteAsync();
}
catch (AIApiException ex)
{
    Debug.LogError($"API Error: {ex.Message}");
}
catch (Exception ex)
{
    Debug.LogError($"Unexpected Error: {ex.Message}");
}

Method Naming Convention

Prefix	Purpose	Examples
GEN	Generate/Create content	GENCompletion, GENImage, GENSpeech
Upload	Upload files	UploadFile, UploadImage, UploadAudio
Download	Download files	DownloadFile
List	List resources	ListModels, ListVoices, ListFiles
Get	Retrieve single resource	GetModel, GetVoice, GetCredits
Delete	Remove resource	DeleteFile, DeleteModel
Count	Count tokens	CountTokens
Tokenize	Get token IDs	Tokenize

Extension Method Hosts

Different types can host different methods:

// String extensions
"text".GENCompletion()
"text".GENResponse()
"text".GENCode()
"text".GENImage()
"text".GENSpeech()
"text".GENEmbed()
"text".Tokenize()
"text".CountTokens()

// Texture2D extensions
texture.GENInpaint("instruction")
texture.ImageToImage("instruction")
texture.GENVideo()

// AudioClip extensions
audioClip.GENTranscript()
audioClip.GENTranslation()
audioClip.GENVoiceChange()
audioClip.GENAudioIsolation()

// Api enum extensions
Api.OpenAI.UploadFile(file)
Api.OpenAI.ListModels()
Api.OpenAI.GetModel("gpt-4o")
Api.OpenAI.GetCredits()

// Model extensions
model.FineTuneModel(trainingFileId)

// Message extensions
message.GENCompletion()

// Prompt extensions
prompt.GENCompletion()
prompt.GENResponse()
prompt.GENCode()

Next Steps

• Request vs Execute - Understanding when API calls happen

• Text Generation - Text generation examples

• Image Generation - Image generation examples