Draggy: A Windows desktop app for local and cloud AI chat and coding (github.com)

Draggy has launched as a free, open-source Windows desktop app designed to serve as a local AI assistant and coding agent. Operating without the need for user accounts or telemetry, Draggy facilitates running AI models, including open models like llama.cpp with GGUF weights from Hugging Face, directly on a user's machine. It allows for coding assistance by reading, editing, and executing commands within project folders, displaying changes through an intuitive diff feature. The app is built using Electron, React, and TypeScript and supports running large models that don’t fit in memory via a secondary engine, colibri, optimizing performance through local storage solutions. For the AI/ML community, Draggy is significant as it enhances accessibility to AI tools by providing a local, privacy-focused alternative to cloud-based solutions. It supports diverse functionalities, including project management, semantic retrieval, customizable interfaces, and voice interactions, along with the ability to generate images using local GPU resources. With separate histories for general chat and coding, interactive task lists, and various permission modes, Draggy empowers users to engage with AI in a more controlled and personalized manner. Its integration with multiple third-party providers elevates its versatility, marking a notable step towards democratizing AI development tools for Windows users.