deepseek-harness-channels
Community channel plugins for DeepSeek Harness: core, visual config, and Feishu channel.
- Stars
- 0
- Language
- JavaScript
- Created
- Aug 28, 2026
- Updated
- Sep 1, 2026
Introduction
DeepSeek Harness Channels
中文文档:README.zh-CN.md
Official and community channel integration packages for DeepSeek Harness (DSH).
Connect external messaging platforms (Feishu/Lark, WeCom, iMessage, etc.) directly to your local DeepSeek AI agents through a unified, zero-relay message bus architecture.
📦 Package Matrix
| Package Name | npm Version | Type | Description |
|---|---|---|---|
@anarkhgatsby/deepseek-harness-core | 0.1.0 | Core Library | Unified message bus, session lifecycle, workspace routing, deduplication, and serial queuing |
@anarkhgatsby/deepseek-harness-channel-config | 0.1.4 | Visual Console | Settings UI for channel management, official Squircle brand assets, Dark Mode adaptive, and parameter configuration |
@anarkhgatsby/deepseek-harness-channel-feishu | 0.1.0 | Channel Adapter | Feishu / Lark AI Bot channel (Official WebSocket outbound long-connection mode) |
@anarkhgatsby/deepseek-harness-channel-wecom | 0.1.1 | Channel Adapter | WeCom (企业微信) AI Bot channel (Official WebSocket long-connection stack) |
@anarkhgatsby/deepseek-harness-channel-imessage | 0.1.1 | Channel Adapter | macOS local-only iMessage channel (Local chat.db listener + AppleScript outbound) |
⚡ Quick Install
Install the desired packages into your DeepSeek Harness web profile:
# 1. Install core message bus and visual config console (Required)
dsh plugin --profile web add @anarkhgatsby/deepseek-harness-core
dsh plugin --profile web add @anarkhgatsby/deepseek-harness-channel-config
# 2. Install channel adapters as needed
dsh plugin --profile web add @anarkhgatsby/deepseek-harness-channel-feishu
dsh plugin --profile web add @anarkhgatsby/deepseek-harness-channel-wecom
dsh plugin --profile web add @anarkhgatsby/deepseek-harness-channel-imessage
Note: Restart DeepSeek Harness after installation to view all channels under Settings ➔ Channels.
📖 Channel Setup Guides
1. 🕊️ Feishu / Lark Setup
The Feishu channel uses official outbound WebSocket long-connection mode. No public IP, webhook URL, or reverse proxy is required.
- Open the Feishu Open Platform Console and create an enterprise custom app.
- Under Add Capabilities, add the Bot capability.
- Under Permissions, grant:
im:message,im:message:send_as_botim:message.group_at_msg:readonly,im:message.p2p_msg:readonly
- Under Events and Callbacks ➔ Event Configuration, select "Use long connection to receive events" and add
im.message.receive_v1. - Create and publish an app version.
- Copy
App IDandApp Secretfrom Credentials & Basic Info, paste them into DeepSeek Harness channel settings, and toggle the channel on.
2. 💬 WeCom (企业微信) Setup
The WeCom channel connects directly to Tencent's cloud WebSocket gateway (wss://openws.work.weixin.qq.com).
- Log in to the WeCom Admin Console.
- Navigate to App Management ➔ AI Bot / Custom Apps and create an AI Bot.
- Retrieve your
botIdandsecret. - In DeepSeek Harness Settings ➔ Channels, configure
botId,secret, and the target workspace, then toggle the channel on.
3. 🍏 iMessage Setup (macOS only)
The iMessage channel operates completely locally. Inbound messages are read from ~/Library/Messages/chat.db and outbound messages are sent via native AppleScript. No message content ever touches a third-party cloud service.
- Ensure macOS Messages.app is signed in with your Apple ID.
- Go to macOS System Settings ➔ Privacy & Security:
- Grant Full Disk Access to DeepSeek Harness.
- Grant Automation permission for DeepSeek Harness to control Messages.app.
- Open DeepSeek Harness Settings ➔ Channels and enable iMessage.
🔒 Privacy & Security
- Zero Cloud Relays: All connections are point-to-point (direct to official WebSocket gateways or local database).
- Local Credentials: API secrets are stored locally in
~/.dsh/settings.yamlwith strict filesystem permissions. - Session Isolation: Each external user/chat is routed to an isolated session space with configurable workspace roots.
📄 License
MIT. See LICENSE.