Agent preset bundle for wish and task management: create wishes or tasks from conversation, opportunity-day check-ins, micro-actions, plus a memory store and progress charts in the web UI.
Install
# from npm (prebuilt)
dsh plugin --profile web add @starwish-ai/xingyuan-dsh
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:starwish-ai/xingyuan-dsh
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
简体中文 | English
XingYuan is a wish & habit-building companion, shipped as a DeepSeek Harness plugin bundle + agent preset.
Features
- Conversation-driven: create wishes / tasks in natural language with duplicate detection
- Opportunity-day check-ins: check-in, make-up, cancel, and future pre-checks; progress always recomputed against today
- Commitment-based progress: only claimed tasks count; a wish achieves when everything committed is done and nothing is left to claim — full progress with unclaimed tasks leaves it "wrapping up" for your call (claim to continue, or delete to achieve), never nagging, never deciding for you
- Micro-actions: break an overwhelming goal into 3–7 tiny steps and walk through them guided
- Growth system: levels Lv.1–Lv.10, streak bonuses, stat cards, 30-day bar chart
- Session view tabs: Today / Wishes / Tasks / Calendar / Growth / Memory — buttons call action endpoints directly; shown only in XingYuan-preset sessions by default, switchable to always show/hide and per-tab in Settings
- Memory: important memories injected into context (limit configurable), full CRUD and clear-all
- Charts: 15 chart types rendered as cards
- Parity between UI and chat: anything you can do on a page — check in, claim, delete, rename or recolor a category, clear memories, adjust preferences — you can also ask the assistant to do, through the same business layer with the same result; loosening the write-confirmation policy stays user-only (Settings page) by design
- Safe writes: write confirmation is per-category (create / check-in / undo / claim / edit / save memory); deletion always confirms and cannot be turned off; defaults match previous versions
- Theming: light/dark follows the app theme
Tools are only mounted on sessions using the XingYuan preset; other sessions are unaffected.
Install
Requires DeepSeek Harness
0.1.7-rc.1or0.2.0-rc.1(peer dependencies admit both version lines, including the host's client-half packages; 0.1.7 rewrote the settings subsystem, so on older hosts this plugin's client half does not activate at all, and since rc.1 dsh refuses to load a whole bundle whose peer range does not match — this package's "every supported line is admitted" claim is locked by a test). Scheduled reminder capabilities differ between the two host versions and are not hardcoded here. Check the upgrade notes in AGENTS.md before bumping dsh).
dsh plugin --profile web add @starwish-ai/xingyuan-dsh
After starting the Web GUI, installation succeeded when "星愿" (XingYuan) appears in the agent picker.
Not in the picker? Upgrade to
0.6.5-alpha.2or later — as of dsh 0.1.7 the host no longer scans~/.dsh/.agent-presets/; presets are registered by a declaration row in the plugin patch, so older releases of this bundle silently never show up.
Data & Backup
Business data lives at ~/.dsh/xingyuan/xingyuan.sqlite and survives uninstall / upgrade; backup is just copying that directory.
Your preferences moved with dsh 0.1.7 into the profile document
(~/.dsh/profiles/<profile>/cordis.patch.yml, this plugin's row), so they are not in
that directory — copy the profile file too if you want them on a new machine.
Settings (Web GUI → Settings → XingYuan)
| Option | Description |
|---|---|
| Coach style | Gentle / humorous / strict |
| User profile | Nickname, occupation, interests |
| Write confirmation | Master switch + per-category toggles (create / check-in / undo / claim / edit / save memory); deletion always confirms and cannot be turned off |
| Memory injection limit | Max memories injected per turn (also settable by saying "use 20 memories" in chat) |
| Tab visibility | Three modes (follow session / always show / always hide) plus per-tab toggles; defaults to follow session. These four non-authorization preferences are also adjustable in chat |
| Confirm card language | Language of the in-chat confirmation card (Chinese / English; the platform does not expose UI language to plugins, defaults to Chinese) |
Development
pnpm install
pnpm build # tsc + tsdown (integration tests depend on lib/)
pnpm test # vitest
Architecture, domain semantics and client-side discipline live in AGENTS.md (Chinese); contributing guide in CONTRIBUTING.
Releasing: push a v* tag to build and publish to npm automatically (prereleases get the alpha dist-tag).
License
Links
More in this category
loopx-project/loopx#dsh-loopx-plugin★ 6233
LoopX, a provider-neutral, local-first state kernel and control plane for long-horizon agents: keeps Goal, Todo, gate, evidence, quota, recovery, and handoff state above DeepSeek Harness, while the plugin bootstraps the CLI and skills, admits bounded same-session continuation, and adds a loopback GoalBar for the exact bound loop.
Q00/ouroboros#integrations/dsh-plugin★ 6196
Config-only bundle that mounts Ouroboros through the DSH MCP client, exposing 36 interview, Seed, execution, evaluation, and evolution workflow tools in DSH.
chuspeeism/dashi-taskboard#deepseek-harness★ 3329
Embeds the active installed Codex Taskboard runtime in the DeepSeek Harness sidebar, using its launcher runtime descriptor instead of a fixed port.
NanmiCoder/dsh-agent-teams★ 2012
AgentTeams multi-agent teams.
EthanYoQ/AI-Novel-Writer#dsh-ai-novel-writer★ 1393
Installs a dedicated AI novel-writing preset and workbench: revisioned local project assets, a compact side drawer, and native approval-gated single-file changes.
tong-io/tongflow#dsh-tongflow★ 1042
TongFlow film-crew studio for image, voice, music and video production: the agent writes per-asset TongFlow workflow files (.tongflow.json) that run through TongFlow plugins, with an embedded workflow canvas, a shot/character/take project layout and a manga-drama template; sessions starting with @tongflow open the Studio view.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.