Custom Prompts & Settings

Fine-tune how the AI processes your hierarchy by editing the system prompts, customizing token budgets, adjusting model temperatures, and adding your studio’s custom naming rules directly into the engine settings.

Accessing AI Custom Settings

All AI configuration rules are managed under Unity’s standard settings hierarchy:

Edit ➔ Project Settings... ➔ AI Scene Organizer

Scroll down to the AI Prompts & Modeling Configuration section.

Essential Configurations

1. Temperature Control

The Temperature slider (range 0.0 to 1.0) controls the randomness of the model:

0.0 - 0.2 (Recommended): Low randomness. Force the model to strictly follow naming structures, casing rules, and group layouts. Use this for deterministic organization.
0.7 - 1.0: High randomness. Use this in the Atmospheric Controller tab to get more creative, artistic lighting ideas and mood profiles.

2. System Instructions Override

The System Instructions text area allows you to define the AI’s persona. The default prompt enforces valid JSON formats and restricts responses to valid operations. You can append your studio rules here:

Example addition: "All 3D static environment mesh folders must start with the prefix 'Static_'. All dynamic characters must go into a folder named 'Entity_Character'."

Token & Prompt Optimization

When organizing very large scenes, the context window can fill up quickly. To reduce API costs and latency, utilize the optimization parameters in the settings:

Filter Static Items: Exclude static meshes from the prompt layout if you only need to group active lights and UI components.
Truncate Deep Trees: Set the Max Scan Depth slider (default 8) to prevent the scanner from parsing extremely deep children, saving token bandwidth.
Exclude Transform Vector Matrices: When enabled, the scanner excludes position coordinates from the payload, sending only GameObject names and component types. This decreases token count by up to 50%.