Responses API

The Responses API provides the most advanced text generation capabilities with support for complex reasoning, tool calling, and multi-modal inputs.

Basic Usage

string response = await "Write a technical specification"
    .GENResponse()
    .SetModel(OpenAIModel.GPT4o)
    .ExecuteAsync();

Input Types

1. String Input

string response = await "Explain quantum mechanics"
    .GENResponse()
    .ExecuteAsync();

2. ConversationItem Input

var userMessage = new UserMessage("Hello, AI!");
string response = await userMessage
    .GENResponse()
    .ExecuteAsync();

3. Prompt Input

var prompt = new Prompt("Explain {topic} in simple terms");
string response = await prompt
    .GENResponse()
    .ExecuteAsync();

Key Features

1. Advanced Reasoning

The Responses API excels at complex reasoning tasks:

string analysis = await @"
Analyze the following code and suggest improvements:
public class Player {
    public int health;
    public void TakeDamage(int amount) {
        health -= amount;
    }
}"
    .GENResponse()
    .SetModel(OpenAIModel.GPT4o)
    .ExecuteAsync();

2. Tool Calling Support

Built-in support for function calling:

var response = await "What's the weather in Tokyo?"
    .GENResponse()
    .SetTools(weatherTools)
    .ExecuteAsync();

// Response may contain tool calls that need to be executed

3. Multi-Modal Input

Support for text, images, and other content types:

var message = new UserMessage();
message.Content.AddText("What's in this image?");
message.Content.AddImage(texture);

string response = await message
    .GENResponse()
    .ExecuteAsync();

Configuration

Model Selection

string response = await "Complex task"
    .GENResponse()
    .SetModel(OpenAIModel.GPT4o)
    .ExecuteAsync();

Temperature & Creativity

string creative = await "Write a story"
    .GENResponse()
    .SetTemperature(1.5f)
    .ExecuteAsync();

string factual = await "What is 2+2?"
    .GENResponse()
    .SetTemperature(0.0f)
    .ExecuteAsync();

Max Tokens

string summary = await "Summarize this long text..."
    .GENResponse()
    .SetMaxTokens(200)
    .ExecuteAsync();

Streaming

Get real-time responses as they're generated:

await "Tell me a long story"
    .GENResponse()
    .StreamAsync(
        onToken: token => Debug.Log(token),
        onComplete: response => Debug.Log("Complete!")
    );

Provider Support

Provider	Support	Models
OpenAI	✅ Full	GPT-4o, GPT-4
Anthropic	✅ Full	Claude 3.5 Sonnet
Google Gemini	⚠️ Partial	Gemini 1.5 Pro
OpenRouter	✅ Full	Various

Note: Not all providers support all Responses API features. Check provider documentation for limitations.

Differences from Chat Completions

Feature	Chat Completions	Responses API
Tool Calling	Basic	Advanced
Multi-turn	Manual	Automatic
Context Window	Standard	Extended
Reasoning	Standard	Enhanced
Complexity	Simple	Advanced

When to Use

✅ Use Responses API for

• Complex reasoning tasks

• Multi-step problem solving

• Tool calling workflows

• Long-form content generation

• Advanced features

❌ Use Chat Completions for

• Simple Q&A

• Basic chatbots

• When maximum provider compatibility is needed

• Cost-sensitive applications (Responses API may be more expensive)

Examples

Example 1: Complex Analysis

async UniTask<string> AnalyzeCode(string code)
{
    return await $@"
Analyze this code for:
1. Performance issues
2. Security vulnerabilities
3. Best practices violations

Code:
{code}
"
        .GENResponse()
        .SetModel(OpenAIModel.GPT4o)
        .SetTemperature(0.3f)
        .ExecuteAsync();
}

Example 2: Multi-Step Reasoning

async UniTask<string> SolveProblem(string problem)
{
    return await $@"
Solve this step by step:
{problem}

Show your reasoning at each step.
"
        .GENResponse()
        .SetModel(OpenAIModel.GPT4o)
        .ExecuteAsync();
}

Example 3: With Context

var conversation = new ConversationItem[]
{
    new SystemMessage("You are a Unity expert"),
    new UserMessage("How do I optimize my game?"),
    new AssistantMessage("Here are some tips..."),
    new UserMessage("Tell me more about object pooling")
};

string response = await conversation.Last()
    .GENResponse()
    .ExecuteAsync();

Best Practices

✅ Do

• Use for complex, multi-step tasks

• Leverage tool calling when available

• Set appropriate max tokens for cost control

• Use streaming for long responses

❌ Don't

• Use for simple tasks (wasteful)

• Ignore provider limitations

• Forget error handling

• Assume all providers support all features

Next Steps

• Code Generation - Specialized code generation

• Structured Output - Type-safe JSON responses

• Chat Completions - Simpler alternative