UAA - Unity AI Assistant
Utilize LLMs to develop in natural language inside Unity engine.
UAA is a tool that integrates Large Language Models (LLMs) with the Unity engine, enabling developers to execute commands using natural language whether you're using open-source models locally with LM Studio or paid models through OpenAI.
- Chat directly with the AI Assistant within Unity, eliminating the need to switch between different platforms (back and forth to ChatGPT, Bing or any other platform out there...)
- The chat has everything you expect from a LLM Chat Bot like saving / loading chat history, automatic naming of the chat...
- Upcoming: Scene-awareness feature to provide context-specific assistance based on the current open scene and project.
- Issue natural language commands to the AI Assistant, which converts them into executable Unity scripts.
- UAA mitigates LLM inconsistencies by checking the console for errors and making corrections automatically.
- Upcoming: Self-validation and user validation features to ensure scripts deliver the desired outcomes.
- Customize the AI Assistant's behavior by adjusting prompts, iteration count, and request settings.
"YoutubeLink.com"
Link to the asset store (coming soon)
Project supports Unity Package Manager. To install the project as a Git package do the following:
- In Unity, open Window -> Package Manager.
- Press the + button, choose "Add package from git URL..."
- Enter "GithubUrl.com" and press Add.
Copy the Assets/Editor/UAA folder to your project's Assets/Editor directory.
- Download LM studio: LM Studio - Discover, download, and run local LLMs
- In LM Studio Download Mistral-7B-Instruct-v0.2-GGUF or any model that you would like to try: TheBloke/Mistral-7B-Instruct-v0.2-GGUF · Hugging Face
- Set up the model settings and the Server configuration and Start Server.
- Head to API keys - OpenAI API
- Create a new secret key - API Key
- in
Assets/Editor/UAA/ProtectedEnv/UAAProtectedEnv.cspast yourOPENAI_API_KEY.
- Dev-Env Awareness: Develop a feature for UAA that understands the current development environment (Scene Object Hierarchy and the project as whole.)
- Project Embeddings: Explore embedding the entire project repository into the model for contextual understanding.
- Train / Fine-Tune a Model: Train the LLM on the Unity API and fine-tune it with command-to-task-to-script examples.
- Safe Execution Checkpoints: Establish a system for saving temporary scene states before script execution.
- User Task Validation: Implement a feature where the LLM provides a task description and script for user validation and feedback.
- Self-Validation: Create a self-validation mechanism instead of User Task Validation.
- Chat History Summarization: Add a function to summarize extensive long chat histories.
- Script Repository: Save successfully executed and validated scripts for future reference.
- UAA Debugger.
- Enhanced Assistant Responses: Improve quality by initializing the assistant response.
- Additional Detail Inquiry: Enable the LLM to request more task details when necessary.
- Prompt Guidelines: Establish guidelines for consistent responses.
- Voice Commands.
- Add amuck display that takes ins consideration things like bold code highlight.
Mehdi Ali : AliMehdi.offi@gmail.com






