basketikun/infinite-canvas

This project is called infinite-canvas. The README, written mainly in Chinese, presents it as an open-source workstation for image creation that puts a free-form canvas, AI image generation, reference-image editing, a chat assistant, a prompt library, and a pile of source images into a single web interface. The intent is to let one person explore visual ideas and iterate on image results without switching between separate apps. The author lists four groups of features. The canvas itself supports multiple projects, draggable and zoomable nodes, connection lines, a small overview map, undo and redo, and import or export. AI creation supports text-to-image, image-to-image, reference-image editing, and text question-and-answer through any service that follows the OpenAI-compatible API shape. A canvas-side assistant can hold a conversation around the selected node and its upstream nodes, generate images, and drop the results back onto the canvas. A prompt library is built up by scraping several GitHub open-source projects and grouping hundreds of image prompts into example sets. The README is up-front about the project's state. It carries a warning that the project is under active development, that database schemas and storage formats may change without compatibility for old data, and that it is more suitable for personal or local deployment than for sharing a public deployment with many users. People who need stability are told to fork the repository and keep their own branch. The technology stack listed in the README uses Next.js with React, TypeScript, Tailwind CSS, Ant Design, Zustand, and TanStack Query for the front end. The back end is written in Go with the Gin web framework and the GORM database library. Deployment is via Docker. Quick start instructions are short. Clone the repository, copy .env.example to .env, edit default account and password values, then run docker compose. The app listens on port 3000 by default. The licence is GNU AGPL v3.