Gomoon is a desktop application designed to enhance productivity and learning by integrating large language models (LLMs). It allows users to create custom AI assistants, interact with various LLMs through a unified interface, and leverage features like continuous conversation, history management, and text manipulation. The tool aims to provide a more accessible and integrated experience for LLM-powered tasks compared to web-based solutions.

How It Works

Gomoon acts as a client for multiple LLM providers, supporting a wide range of models including OpenAI (GPT-4o, GPT-4 Mini), Wenxin, DeepSeek, Qwen, Gemini, Kimi, Claude, and Llama/Ollama via node-llama-cpp. Users configure their chosen model engines within the application. Key features include quick invocation via hotkeys, text selection integration for context passing, file/URL parsing, web search, text-to-speech, and a "memory capsule" for local knowledge bases.

Quick Start & Requirements

• Installation: Download from the official website. Mac users may need to right-click and open to bypass security warnings.
• Prerequisites: For node-llama-cpp with CUDA support: NVIDIA GPU, CUDA Toolkit (v12+), Node.js (v20.11.1+), Yarn (v1.2.22+).
• Setup: For local development with CUDA, clone the repo, install dependencies (yarn), download CUDA support (npx --no node-llama-cpp download --cuda), and potentially copy resource files.
• Resources: Official download: https://gomoon.ai/

Highlighted Details

• Supports over a dozen LLM providers and custom OpenAI-compatible endpoints.
• Features "memory capsules" for secure, local knowledge storage.
• Enables quick text selection, pasting, and context-aware actions via hotkeys.
• Offers assistant import/export for sharing and community collaboration.

Maintenance & Community

• Active development with community contributions noted.
• QQ group available for discussion and updates: 758015092.

Licensing & Compatibility

• The README does not explicitly state a license. Compatibility for commercial or closed-source use is undetermined.

Limitations & Caveats

• Mac users require manual permission granting and app restarts for certain features like double-click copy and file sending.
• Local model resource files are not included in the default GitHub repository and must be manually downloaded or copied from an installed instance.
• The lack of a specified license raises concerns for commercial adoption.