Text Generation

Text generation allows you to create natural language output from a prompt using powerful large language models (LLMs) like GPT-4o, Gemini, and more.

This is commonly used for answering questions, summarizing content, writing marketing copy, or even creative writing.

✅ Basic Usage

Use the Fluent API to quickly create and execute a text generation task.

string prompt = "Explain quantum computing in simple terms.";

string result = await prompt
    .GENText()
    .SetModel(GPTModel.GPT4o)
    .SetTemperature(0.7f)
    .ExecuteAsync();

Debug.Log(result);

💡 GENText() is an extension method for string, turning your prompt into a text generation task.

---

⚙️ Available Configuration

You can customize the task using the following fluent methods:

Method	Description
SetModel(model)	Choose the model to use (e.g., GPTModel.GPT4o, GeminiModel.Gemini1_5Pro).
SetInstruction(text)	Set a system-level instruction (defines tone/role).
SetStartingMessage(text)	Set the first assistant message to guide the response.
SetMaxOutputTokens(int)	Limit the maximum output length (in tokens).
SetTemperature(float)	Control randomness: 0.0 is focused, 1.0 is creative.
SetTopP(float)	Nucleus sampling: limit token selection to top probability P.
SetTopK(float)	Limit selection to the top K tokens (used by Gemini).
SetSeed(int)	Use a fixed seed for consistent output.
SetPresencePenalty(float)	Discourage repeated ideas.
SetFrequencyPenalty(float)	Discourage repeated words.

---

🚀 Streaming Output

For real-time applications like chat or narration, you can stream the output word-by-word:

await prompt
    .GENText()
    .SetModel(GPTModel.GPT4o)
    .StreamAsync(
        onTextReceived: chunk => Debug.Log(chunk),
        onStreamStart: () => Debug.Log("Started..."),
        onStreamEnd: () => Debug.Log("Finished.")
    );

✅ Supported only on models that support streaming (e.g., OpenAI GPT family).

---

📦 Example Result

Quantum computing is a type of computing that uses quantum bits, or qubits, which can represent both 0 and 1 at the same time...

---

🧠 Tips

• You don’t need to manually handle UniTask cancellation unless you want to cancel early.

• The default model is defined in your project settings (AIDevKitSettings.DefaultTextModel).

• You can safely combine this with object generation if you want to extract structured data — see C# Object Generation.