1
stars
0
forks
MIT
License
2026-08-15
Created
2026-08-15
Last push
README
dsh-weixin-bridge
Chat with your local DeepSeek Harness (DSH) from WeChat. Send a message on your phone; DSH runs it as a real task on your computer (tools, files, commands); the reply comes back to WeChat.
No public IP needed, no desktop WeChat, nothing to install on your phone — it talks through the official WeChat iLink channel (the same API WeChat's own bots use) with a QR-code login.
No API key needed. Replies are produced by your local DSH in agent mode — it reuses the model already configured in DSH. You never paste a DeepSeek/OpenAI key into this tool.
How it works
WeChat message
→ bridge polls the official iLink channel (long-polling)
→ task runs through `dsh --profile headless "message"` (a DSH dialog / agent on your PC)
→ the agent's reply is sent back to WeChat
- Context kept: the bridge remembers the last 20 turns, so the bot keeps one continuous dialog instead of forgetting everything each message.
- Progress feedback: you immediately get "⏳ processing…" and periodic updates while a task runs — no silent waiting.
- Only you: only the WeChat account that scanned the QR code can trigger tasks.
Installation guide (from zero)
- Install Node.js >= 20 — https://nodejs.org
- Install DeepSeek Harness (global, so its
bin.jslives under Node'snode_modules):
Then verify the one-shot mode works:npm install -g @deepseek-ai/dshdsh --profile headless "hi" # should print a reply; if it errors, check the headless profile / your DSH installIf DSH is installed elsewhere, set
DSH_BINto itsdsh/lib/bin.jspath before starting the bridge. - Install this tool:
npm install -g dsh-weixin-bridge # or clone https://github.com/Culeot/dsh-weixin-bridge and npm install - Log in with WeChat (one time):
node login.js # a QR code appears → scan with WeChat → confirm → the bot is linked - Start the bridge:
node bridge.js # logs "bridge started, waiting for WeChat messages" - Test: in WeChat, message the bot — "hi", or "list what's in C:\ai项目". You should get "⏳ processing…" then the reply.
Configuration
Edit the top of bridge.js:
| Constant | Default | Meaning |
|---|---|---|
AGENT_CWD |
C:/ai项目 |
working directory for agent tasks (change to your own) |
AGENT_TIMEOUT |
300s | per-task timeout |
POLL_INTERVAL |
5s | WeChat poll interval |
MAX_HISTORY |
20 | conversation turns kept for context |
The bot token is saved to account.json after login (git-ignored).
Run at startup (optional)
Register in the registry:
HKCU\Software\Microsoft\Windows\CurrentVersion\Run
WeixinBridge = powershell -WindowStyle Hidden -Command "Start-Process node -ArgumentList 'C:\path\to\bridge.js' -WindowStyle Hidden"
Troubleshooting
- Bot doesn't reply: check
bridge.login the project folder; most issues aredsh --profile headlessfailing — run that command manually first. - "找不到 DSH / DSH not found": DSH isn't reachable from this Node — install it globally or set
DSH_BIN. - QR code expired: re-run
node login.js. - Slow replies: headless starts a fresh DSH process per message (cold start takes a few seconds); progress messages tell you it's working.
- Model config: the agent uses whatever model DSH is configured with (set it in DSH, e.g.
deepseek-v4-flash). No key needed here.
Security notes
- Only the account that scanned the QR code can trigger tasks — nobody else's messages are acted on.
- A task can execute commands on the PC. Keep the token private:
account.jsonandbridge.logare git-ignored. - Uses the official WeChat iLink API; no third-party WeChat protocol involved.
Files
| File | Purpose |
|---|---|
bridge.js |
the bridge: poll → headless task → reply (with context & progress) |
login.js |
QR-code login, saves the bot account |
License
MIT
More in Chat & IM
voyager
by nagi-ovo
Enhancement suite for Gemini, AI Studio, Claude & ChatGPT — plus a prompt manager for any web UI, DeepSeek Harness included. / 面向 Gemini、AI Studio、Claude 与 ChatGPT 的增强套件;提示词管理器可用于任意 Web UI,含 DeepSeek Harness。
awesome-gpt-image-2
by freestylefly
Prompt as Code | GPT-Image2 工业级提示词引擎与模板库,470+ 个案例逆向工程,20+ 套工业级模板,并提炼出Skills,持续更新中
seewxapkg
by leslie-sss
极简实用的微信小程序反编译 Web 工具
dsh-lark-bot
by plutokeating
dsh-lark-bot:把 DeepSeek Harness (dsh) 桥接进飞书/Lark 的 bot:流式卡片、项目工作区、并行任务、多角色 Agent、跨会话通知、对话内模型/密钥管理与安全网守护(dsh 崩溃后飞书仍可自救)。A bridge bot connecting DeepSeek Harness (dsh) into Feishu/Lark: streaming cards, workspaces, parallel tasks, multi-role agents, cross-session notify, in-chat model/key management, and a safety-net guardian.
