Google Gemini Configuration

Step-by-step guide to connect Google’s Gemini models for rapid environment generation

Google Gemini is the default recommended provider for AI Level Scatter. It offers low latency, high token limits, and robust text-to-JSON formatting capabilities, making it ideal for compiling layout profiles and providing real-time design advice in the Editor.

Step 1: Obtain a Gemini API Key

To connect the Unity Editor to Google’s servers, you must have a valid developer API key:

Visit the Google AI Studio Console.
Sign in with your Google account.
Click the Get API Key button in the sidebar.
Select Create API Key in new project (or choose an existing Google Cloud project).
Copy the generated API key string.

Step 2: Configure Unity Editor Settings

Once you have your key, assign it inside the Unity Editor:

Open Unity and select Tools → Decnet → AI Settings Manager from the top menu.
Select Google Gemini from the Active Provider dropdown.
Paste your copied API key into the API Key input field.
In the Model Name dropdown, select your preferred model:
- gemini-1.5-flash: Recommended for high-speed prompt generation and low-latency chat queries.
- gemini-1.5-pro: Recommended for complex 3D room generation and detailed layout analyses.
Click Save Settings.

Step 3: Verify the Connection

Click the Test Connection button at the bottom of the Settings Manager. The engine will send a lightweight ping request to Google’s API endpoint: https://generativelanguage.googleapis.com/v1beta/models/

Successful Connection: The status indicator turns green, showing your connection latency (e.g. Latency: 110ms (OK)).
Failed Connection (Unauthorized): If you receive a 403 Forbidden or 401 Unauthorized error, verify that your API key is correct and has not expired, and that Google AI Studio services are active in your region.

Advanced API Parameters

The Gemini provider configures request structures behind the scenes to optimize generation stability:

{
  "contents": [{ "parts":[{ "text": "YOUR_PROMPT" }] }],
  "generationConfig": {
    "response_mime_type": "application/json",
    "temperature": 0.2
  }
}

JSON Mode: Decnet automatically requests the application/json mime-type for prompt presets. This forces Gemini to return structured placement coordinates rather than standard conversational text.
Low Temperature: The temperature is locked at 0.2 to minimize “hallucinated” layout parameters, ensuring that the model strictly generates valid float scales and integer densities.

[!NOTE] Google Gemini provides a free tier with rate limits suitable for development. If you hit rate limits during intense prompt generation sessions, consider upgrading to the pay-as-you-go tier in Google AI Studio to increase your requests per minute.