About this project

StudyMate is an open-source AI learning companion project, positioned as a "math/computer science learning workflow, SKILL, and HTML courseware engine," with the core philosophy of "learn with doing." **Multi-platform support**: The project can run as a "learning mode" preset for DeepSeek Harness (DSH), be packaged as a native multi-agent plugin for Google Antigravity, or be imported into Codex / ChatGPT Work via a ZIP package. DSH requires Node.js 22/24+, Python 3.9+, and PyYAML. **Five-role agent division of labor**: The system organizes five roles as needed—resource-scout (gathering materials from authoritative textbooks and official documentation), image-scout (fetching web images to build an image pool), curriculum-designer (arranging a knowledge point roadmap by prerequisite dependencies), learning-coach (writing courseware explanation content), and practice-evaluator (the sole owner of the question bank). When the host supports it, tasks can be delegated to sub-agents; otherwise, they are completed sequentially. **Course structure**: Each course has one outline, with knowledge points arranged into a roadmap layered by dependencies, and each node marked with a lesson type (concept lesson/hands-on lesson/lab lesson). Courseware is rendered in HTML, styled close to classic textbooks, with rich illustrations, customized questions, and project goals. Learning progress is written to files and can be inherited in new conversations. **Cross-subject shared memory**: The system remembers the user's current level, effective teaching methods, and common sticking points, so there is no need to reintroduce yourself when switching subjects. **Engineering capabilities**: The repository includes 12 skill definitions (master control + 5 roles + 6 protocols), 5 data structure schemas, page templates and front-end assets, homepage generation and courseware rendering scripts, four validation checks (outline/courseware/image pool/skill frontmatter), and npm test regression tests. The example workspace contains two complete courses, Linear Algebra and Computer Networks, which can be browsed immediately after cloning. The project uses the MIT license, is maintained by Cattofu, and offers a QQ discussion group and a long-term maintenance commitment.