top of page
logo
1. Quick Start Guide

To begin using the AI Game Design Architect within your Unity project, follow these quick setup steps:


Open the Settings Provider

  1. Go to Edit-> Project Settings -> AI Design Architect (or Project/AI Design Architect).

  2. Configure Inference Endpoint

    • Endpoint URL: Provide your LLM completion URL. (e.g., Local Ollama, LM Studio, or your custom Cloud/Web API). Defaults to local Ollama: http://localhost:11434/v1/chat/completions.

    • Model Name: Specify the target model (e.g., llama3, mistral).

    • API Key: Leave empty if using local Ollama(or LM Stdio), or enter your secure key/bearer token if using commercial or corporate custom endpoints.

  3. Launch the Window Navigate to the top menu
    Tools -> AI Game Design Architect to open the interactive editor workspace.

​

Need help setting up local servers or corporate proxies? Check out our Detailed Connection Guide (Ollama / LM Studio / Custom Web API) for step-by-step instructions.

2. Dynamic Architecture & Handcrafted Parser

This tool utilizes a robust, fallback-safe manual scanning engine designed specifically for unstable LLM data streams. It operates under strict runtime rules to maintain extreme editor stability:

  • Backslash Parity Check: String extraction utilizes character-parity loops (backslashCount % 2 == 0) to correctly differentiate between standard inner string markdown quotes (\") and actual end-of-value syntax properties.

  • Intelligent Pointer Increments: The parsing loop automatically steps right past raw token limits and stray commas, eliminating the risk of editor lockups or infinite loops if the connection breaks or outputs truncated streams.

3. Advanced Customization (JSON Templates)

The entire design loop and localization framework are driven by external JSON configurations. You can fully tailor the template sections, prompt behaviors, and editor texts.

The tool manages master files safely. On the first initialization, it copies default templates from your asset directory to the local project's secure UserSettings/ directory. This pattern ensures that your custom templates and configurations are never overwritten when updating the asset via the Unity Asset Store.

​

📄 Locale Configuration Template (AIDesignLocale_EN.json)

Allows you to override the interface terminology, welcome prompts, and structural instructions passed to the AI Layer.

📄 Blueprint Template Configuration (AIDesignTemplate_EN.json)

Defines the document blocks that the AI will programmatically extract and populate during your interactive design interview.

4. Enterprise Privacy & Sovereign Security

Because the asset establishes direct socket pipelines from your Unity editor instance straight to your user-configured URL endpoints, DevForge AI never captures, logs, or intercepts your game mechanics, source logic, or text prompts. When utilizing localized inference architectures (e.g., local Ollama), your entire creative workflow remains fully insulated on your workstation or corporate intranet, providing bulletproof defense against intellectual property leaks or cross-border geopolitical data harvesting.

5. Detailed Connection Guide

This section provides step-by-step instructions for configuring your inference endpoints, ensuring smooth communication between the Unity Editor and your AI backends.

5.1. Connecting to Local Ollama
Ollama is a fantastic choice for running models locally. However, by default, it may block API requests originating from other applications (like the Unity Editor) due to Cross-Origin Resource Sharing (CORS) restrictions.

Setup Steps:
1. Ensure Ollama is installed and running on your local machine.
2. To allow Unity to communicate with Ollama, you must set the environment variable to allow all origins before launching the application:

   Windows (Command Prompt):
     set OLLAMA_ORIGINS=*
     ollama serve
  
   macOS / Linux (Terminal):
     OLLAMA_ORIGINS="*" ollama serve

3. In Unity's Project Settings, set the **Endpoint URL** to:
   `http://127.0.0.1:11434/v1/chat/completions`


5.2. Connecting to LM Studio
LM Studio provides a user-friendly interface to manage and run local LLMs. It exposes an OpenAI-compatible API out of the box.

Setup Steps:
1. Open LM Studio and download your preferred model (e.g., `Llama 3.1 8B`, `Mistral`).
2. Navigate to the **Local Server** tab (the developer/server icon on the left vertical menu).
3. Select your model from the top dropdown and click the **Start Server** button. 
4. Take note of the port number provided by the UI (the default is usually `1234`).
5. In Unity's Project Settings, set the **Endpoint URL** to:
   `http://127.0.0.1:1234/v1/chat/completions`


5.3. Connecting to Custom Web APIs / Cloud Services (e.g., Google Gemini)

If you are working within a restricted enterprise environment or prefer cloud-based models, you can route requests through your own cloud infrastructure, internal proxy servers, or commercial API endpoints like Google Gemini.

Since the asset is built with flexible OpenAI-compatibility, connecting to commercial cloud APIs is seamless.


Example: Setting up Google Gemini API (via Google AI Studio)

  1. Obtain an API key from Google AI Studio.

  2. In Unity's Project Settings, configure the following values:

    • Endpoint URL: https://generativelanguage.googleapis.com/v1beta/openai/chat/completions

    • Model Name: gemini-3.1-flash-lite (Or any OpenAI-compatible Gemini model version)

    • API Key: [Your Gemini API Key]

Architecture Note for Enterprise Users: Since AI Design Architect relies exclusively on Unity's native network architecture (UnityWebRequest) and standard C# libraries, it seamlessly inherits your machine's system proxy settings. No external CDNs or Node.js runtimes are required, ensuring 100% compliance with corporate security and network protocols.

bottom of page