jingyunstudio/jingyun-dsh

Jingyun DSH Client – A desktop front‑end for turning AI agents into sellable products

What it is – A Tauri‑based desktop application (and a reusable npm package) that sits on top of DeepSeek Harness (DSH) and the Jingyun Studio SaaS platform. It adds a full commercial loop around any DSH‑powered AI agent: user authentication, subscription / pay‑per‑use billing, token‑quota control, a marketplace for agents, skills and plugins, and cloud sync of data and artefacts.

Key capabilities

Feature	What you get
Member login & subscription	Phone‑number login, support for recurring subscriptions, usage‑based billing and reseller models.
Compute quota management	Built‑in token‑budget system that lets users recharge and limits AI usage to control costs.
Agent library	Identity system for AI agents; each conversation runs in an isolated “agent” context so the same client can switch between many specialised bots.
Skill library	Re‑usable skill packs (e.g., Tencent IMA knowledge base, DingTalk, WeCom, Feishu connectors) that extend an agent’s abilities.
Plugin marketplace	A curated list of 800+ community‑contributed DSH plugins; one‑click install/uninstall from the UI.
App marketplace	Publish your own agents, skills or plugins, set a price, and sell them through the Jingyun Studio store.
Cloud data & asset store	Sync chat logs, knowledge bases and generated artefacts (code, docs, charts) to the cloud.
Product workspace	Drag‑resize side panel that previews and manages multi‑type outputs (code, documents, visualisations).

How you can use it

1. As an npm plug‑in – If you already run a DSH server, install @jingyun-ai/jingyun-dsh and point it at your Jingyun Studio tenant via a small JSON config. The plug‑in injects the full membership, payment and marketplace UI into your existing DSH UI.
2. As a standalone desktop client – Clone the repo, run the build scripts, and use the provided Windows‑only GUI packager (run_pack.bat) to brand the app (name, logo, domain) and produce an installer. The resulting executable ships a Tauri‑Rust shell, a bundled Node/Python/Git runtime, and the UI built with React + TypeScript.

Architecture at a glance

jingyun_dsh/
├─ packages/jingyun-dsh/          # TypeScript UI plug‑in (React components, routes, config)
│   ├─ src/client/               # UI pages & reusable components
│   ├─ src/agent/                # Backend logic for managing AI agent packages
│   └─ resources/                # Built‑in skill bundles, plugin registry snapshot
├─ src-tauri/                     # Tauri v2 native shell (Rust)
│   └─ src/lib.rs                # Main process – window, tray, lifecycle
├─ scripts/                       # Build & packaging helpers (Python/Tkinter GUI, JS helpers)
└─ run_pack.bat                   # One‑click Windows packager

Tech stack

• Desktop shell – Tauri v2 + Rust (native window, system tray, secure API surface)
• AI runtime – DeepSeek Harness (plug‑in‑able LLM orchestration)
• Commercial backend – Jingyun Studio SaaS (membership, payment, quota, marketplace APIs)
• Front‑end – React + TypeScript (distributed as an npm package)
• Monorepo tooling – pnpm + tsdown for dependency management and compilation
• Packaging UI – Python + Tkinter GUI that lets non‑technical teams customise branding before building the installer
• Installer – NSIS (Windows)

Getting started

1. Prerequisites – Node ≥ 18, pnpm ≥ 8, Rust stable (for Tauri builds), optional Python ≥ 3.8 (only for the GUI packager).
2. npm‑plug‑in path

   npm install @jingyun-ai/jingyun-dsh
   # copy example config, fill in your Jingyun domain, then enable the plug‑in in DSH
   

3. Full client path

   git clone <repo>
   cd jingyun_dsh
   pnpm install
   pnpm build            # builds the UI plug‑in
   pnpm tauri:dev        # run Tauri in dev mode (web UI at http://localhost:3080)
   pnpm tauri:build      # produce a production installer
   # or use the GUI packager:
   run_pack.bat
   

4. Branding – Edit packages/jingyun-dsh/desktop-config.json (or use the GUI) to set custom_name, custom_logo and the Jingyun tenant domain.

Who it’s for

• AI startup teams that want to expose their custom agents as commercial products without building a billing system from scratch.
• Developers already using DeepSeek Harness who need a ready‑made UI layer for user accounts, quota management and a marketplace.
• Product managers who want a Windows desktop client that can be white‑labelled for their brand.

License – Apache 2.0 (project) plus the original licenses of DeepSeek Harness, the built‑in skill packs and the community plugin registry.

---

All details above are taken directly from the repository’s README.