Parameters

Configure model parameters and runtime settings.

Overview

Parameters control how the AI model generates responses:

Model - Which AI model to use
Temperature - Randomness/creativity
Max Tokens - Response length limit
Top P - Nucleus sampling
Tools - Available functions
Voice - TTS voice selection

Models

Set Model

// At initialization
settings.Model = "gpt-4o";

// At runtime
agent.Model = "gpt-4o-mini";

Available Models

// OpenAI
agent.Model = "gpt-4o";
agent.Model = "gpt-4o-mini";
agent.Model = "o1";
agent.Model = "o1-mini";

// Anthropic
agent.Model = "claude-3-5-sonnet-20241022";
agent.Model = "claude-3-5-haiku-20241022";

// Google
agent.Model = "gemini-2.0-flash-exp";
agent.Model = "gemini-1.5-pro";

Model Properties

var model = agent.Model;
Debug.Log($"Name: {model.Name}");
Debug.Log($"Context Window: {model.ContextWindow}");
Debug.Log($"Max Output: {model.MaxOutputTokens}");
Debug.Log($"Supports Tools: {model.SupportsTools}");
Debug.Log($"Supports Vision: {model.SupportsVision}");

Temperature

Controls randomness (0.0 - 2.0):

// Low temperature (deterministic, focused)
agent.Temperature = 0.2f;

// Medium temperature (balanced)
agent.Temperature = 0.7f;

// High temperature (creative, random)
agent.Temperature = 1.5f;

Guidelines:

0.0 - 0.3 - Factual, consistent, predictable
0.4 - 0.8 - Balanced creativity
0.9 - 2.0 - Very creative, unpredictable

Max Tokens

Limit response length:

// Short responses
agent.MaxTokens = 500;

// Medium responses
agent.MaxTokens = 2000;

// Long responses
agent.MaxTokens = 4000;

Note: 1 token ≈ 0.75 words

Top P

Nucleus sampling (0.0 - 1.0):

// Very focused
agent.TopP = 0.1f;

// Balanced (default)
agent.TopP = 1.0f;

Usually leave at 1.0 and control via temperature instead.

Voice

For text-to-speech:

agent.Voice = "alloy";   // Neutral
agent.Voice = "echo";    // Warm
agent.Voice = "fable";   // British
agent.Voice = "onyx";    // Deep
agent.Voice = "nova";    // Friendly
agent.Voice = "shimmer"; // Energetic

Tools & Tool Choice

Add Tools

agent.RegisterToolExecutor(new WeatherTool());

Tool Choice

// Let agent decide
agent.ToolChoice = ToolChoice.Auto;

// Require tool use
agent.ToolChoice = ToolChoice.Required;

// Disable tools
agent.ToolChoice = ToolChoice.None;

// Force specific tool
agent.ToolChoice = ToolChoice.Function("get_weather");

Complete Example

using UnityEngine;
using Glitch9.AIDevKit.Agents;

public class ParameterController : MonoBehaviour
{
    [SerializeField] private AgentBehaviour agent;
    
    // Preset configurations
    public void SetFastMode()
    {
        agent.Model = "gpt-4o-mini";
        agent.Temperature = 0.3f;
        agent.MaxTokens = 500;
    }
    
    public void SetCreativeMode()
    {
        agent.Model = "gpt-4o";
        agent.Temperature = 1.2f;
        agent.MaxTokens = 2000;
    }
    
    public void SetAnalyticalMode()
    {
        agent.Model = "o1";
        agent.Temperature = 1.0f;
        agent.MaxTokens = 4000;
    }
    
    // Dynamic adjustment
    public void AdjustTemperature(float value)
    {
        agent.Temperature = value;
        Debug.Log($"Temperature: {value}");
    }
    
    public void AdjustMaxTokens(int value)
    {
        agent.MaxTokens = value;
        Debug.Log($"Max Tokens: {value}");
    }
}

Preferences Store

Save user preferences:

// Preferences are auto-saved
agent.Model = "gpt-4o";
agent.Temperature = 0.8f;

// Restored on next initialization

Next Steps

PreviousCanceling Requests NextModels

Last updated 17 minutes ago