A campus AI assistant developed by Zhejiang University students that integrates the Academic Affairs Website, Graduate Academic Affairs Website, Learn at ZJU, Smart Cloud Classroom, Chaoxing Learning, PTA, school and college notifications, ZJU email, CC98, sports and arts venues, as well as selected group chats from local WeChat / DingTalk. It organizes this information into personal profiles by course; the sidebar adds three pages: "Schedule," "Campus Affairs," and "Learning." When you ask a question, it queries and compares data across these systems. Originally an Electron desktop version based on secondary development of DSH, 11-02 was changed to a plugin published to npm.
Star count 69
Reasons
- First inclusion in this column; the npm package dsh-zju-assistant released versions from 0.1.0 to 0.1.2 on 11-02. The author states it has been tested on DeepSeek Harness 0.2.1-rc.5. Installation is done by entering the package name on the plugin page, and ten sub-plugins can be toggled individually.
- Actions that change account status (venue reservations, course selection, submitting assignments, second-class activity registration, sending emails) are only executed when explicitly requested and require approval from dsh; a complete draft is provided before sending an email.
- This is a specific example of "the community replacing self-built desktop shells with official plugin pages": the author notes that windows, system tray, auto-updates, and signing no longer need to be maintained personally, allowing the repository to focus solely on campus features.
Usage
- Install the official DeepSeek Harness desktop version (0.2.1-rc.3 or newer) or use npx @deepseek-ai/dsh web; go to "Plugins" in the sidebar → "Add Plugin," enter dsh-zju-assistant, and after enabling it, fill in your Unified Identity Authentication and WebVPN accounts under "Campus Accounts."
- To read WeChat Official Account articles or pass PTA CAPTCHAs, also install dsh-builtin-browser; users of command-line assistants (Codex, Claude Code) can alternatively use the author's other repository zju-skill-public, which uses the same underlying commands.
Risks
- Useful only for Zhejiang University students; this is a personal student project, not an official school product. Credentials such as Unified Identity Authentication passwords are stored locally (the author states they are encrypted by the system credential store), and retrieved campus data is sent to the model service you select in dsh along with the conversation.
- Reading WeChat chats on macOS requires accessing the running WeChat process; this often fails if System Integrity Protection (SIP) is enabled. The author does not own a Mac and has only tested with synthetic data. Licensed under GPL-3.0-only, which requires providing source code for distributed derivative programs.