This project provides a Live2D "Kanban娘" (desktop assistant) application for Windows and Linux, offering features like AI chat, web search, and schedule reminders. It targets desktop users seeking a customizable and interactive virtual assistant experience, enhancing productivity and personalization.

How It Works

Built with Electron, the application leverages a modular design to integrate various functionalities. It supports both cloud-based and local Live2D model loading (V2-V5), with customizable positioning and scaling. AI chat is powered by the ChatGPT API, requiring a user-provided API key. Schedule reminders utilize native notifications, and a "mini-window" mode allows for a less intrusive presence.

Quick Start & Requirements

• Install: Download from the Releases page. An AUR package is available for Arch Linux.
• Prerequisites: Windows or Linux. Mac users need to compile locally. ChatGPT integration requires an OpenAI API key and a stable internet connection. Local model loading may require specific JSON configurations and the Cubism SDK renderer for offline use.
• Setup: Estimated setup time is minimal for pre-compiled binaries. Local compilation requires setting up the Electron build environment.

Highlighted Details

• Supports Live2D V2-V5 models with custom scaling and positioning.
• Integrates ChatGPT API for AI-powered conversations.
• Features native notification-based schedule reminders with customizable sounds.
• Offers a "mini-window" mode and customizable bubble backgrounds.
• Includes a global settings window built with WinUI3 for Windows.

Maintenance & Community

The project is actively developed with ongoing updates and a public project board for tracking progress. Users can contribute via issues and pull requests. Community support channels are not explicitly mentioned, but GitHub Issues serve as the primary interaction point.

Licensing & Compatibility

The project is licensed under GPL v3.0. It incorporates MIT-licensed components (ChatGPT-API). The use of Live2D models is subject to the Cubism EULA. While designed for cross-platform use, Linux support is noted as incomplete, and Mac users must compile from source. Commercial use or linking with closed-source applications may be restricted by the GPL v3.0 license.

Limitations & Caveats

The Tencent NLP service previously used is discontinued. Local model loading is experimental and may have bugs or missing features. Performance overhead and higher power consumption are noted due to the Electron framework, making it less ideal for laptops. The project does not provide API keys and disclaims responsibility for account issues arising from their use.