dsh-im
Connect nine IM bot channels to DeepSeek Harness and continue sessions, switch models, manage workspaces, and handle tool approvals from Feishu, WeChat, DingTalk, Slack, Telegram, and other chat clients.
Project overview
dsh-im is a DSH plugin for DeepSeek Harness Web. It brings Feishu, WeChat, DingTalk, WeCom, QQ, Slack, Telegram, Discord, and WhatsApp into one settings page, where users can continue Harness sessions, switch models and workspaces, handle tool approvals, and optionally connect a local Harness outward to a public service through the AI Office Connector.
Core capabilities
Nine IM channels in one plugin
Manage Feishu, WeChat, DingTalk, WeCom, QQ, Slack, Telegram, Discord, and WhatsApp from one plugin page. Each platform uses QR authorization, a Manifest, or bot credentials as documented.
Remote Harness session control
Use commands such as `/new`, `/stop`, `/steer`, `/compact`, `/models`, `/workspace`, and `/session` to manage the task and session bound to the current chat.
Images and streaming replies
All nine channels can forward supported image formats with optional text to Harness, while replies use platform-specific thinking states, progress updates, streaming cards, or message edits.
Outbound AI Office connection
A local Harness can connect outward to a public AI Office without a public IP or port forwarding, including round trips for tool approvals and follow-up questions.
Installation and usage
Install the stable npm release into the DeepSeek Harness web profile, restart the Web service, authorize one IM channel, and use the bot's `/status` command to verify the connection.
Let an AI Agent install it
Send this prompt to Codex, Claude Code, or another AI agent that can work with your local environment.
Help me install and configure the dsh-im DSH plugin.
Project URL: https://github.com/xmanrui/dsh-im
Read the repository README, package.json, and LICENSE first. Confirm the active DeepSeek Harness Web profile, the Node.js version, and that the dsh command works. Prefer the documented stable command `dsh plugin --profile web add -w @xmanrui/dsh-im`, then restart `dsh web`. Show me how to open Settings > Plugins > IM Bots and help configure the platform I select by QR authorization, App Manifest, or bot credentials. Credentials must only be submitted to the local Harness Host and must not be written to ordinary configuration files, chat logs, or command logs. After setup, confirm that the bot is online and send `/status` to verify the connection. Report the commands, changed paths, and verification result. Ask before requesting any token, opening LAN access, expanding permissions, overwriting files, or approving dependency build scripts.- DeepSeek Harness Web is installed and starts successfully
- Node.js 22.19 or later
- Permission to run dsh commands and modify the web profile
- A QR-capable account, App Manifest, or bot credentials for the selected platform
dsh plugin --profile web add -w @xmanrui/dsh-im- 1Check DSH Web and Node.js
Confirm that DeepSeek Harness Web starts normally, Node.js is version 22.19 or later, and the target profile is `web`. Prepare an IM account or bot application that you control.
- 2Install the stable npm package
Run the documented command. `dsh plugin --profile web add -w @xmanrui/dsh-im` Restart `dsh web` after the command finishes so the web profile loads the plugin.
- 3Open the IM bot settings
Open DeepSeek Harness Web, then go to Settings > Plugins > IM Bots. Choose a platform and follow its flow to scan a QR code, use an App Manifest, or enter the required bot credentials.
- 4Set the workspace and access scope
Review the Harness workspace shown on the bot card. Limit direct messages, group mentions, and bot visibility to trusted users. Telegram can use per-bot private-chat allowlist mode when needed.
- 5Verify the connection and commands
Wait until the bot is online, send `/status` to check the Harness connection, and send `/help` to list commands. Finish with one ordinary message to create a minimal session and confirm that a reply returns.
After restarting `dsh web`, open the plugin settings and confirm that the IM Bots entry is visible. Authorize one channel and wait until the bot is online, then send `/status`. The basic setup is complete when the bot reports that it can connect to DeepSeek Harness.
- Prefer the stable npm release; GitHub source installation fetches and builds a Git dependency
- Limit bot visibility to trusted users because workspace and session commands can expose local paths and session metadata
- Enable Telegram's private-chat allowlist mode when stricter access is required
- Do not put platform secrets, bot tokens, or an AI Office Device Token in ordinary configuration files
Use cases
Continue development tasks from mobile chat
Away from the workstation, a trusted IM bot can report status, steer or stop running work, and continue an existing Harness session.
Connect team chat platforms to one Harness
Teams using Feishu, DingTalk, Slack, or Discord can keep separate workspaces and session bindings for different channels and bots.
Handle tool approvals and follow-up questions remotely
When Harness requests approval or additional information, the interaction can continue in the chat client or the AI Office operator panel.
Connect a local Harness to a public Office
The AI Office Connector provides an outbound path when exposing a public IP, forwarding a port, or operating an inbound WebSocket is undesirable.
Assessment
This assessment is based on the repository README files, package.json, license, and maintenance metadata. dsh-im combines nine familiar IM channels under one settings entry and carries sessions, models, workspaces, images, and human approvals into chat. The installation command is short, and the repository documents credential storage and remote-command boundaries clearly. Bot visibility still requires careful control because workspace lists, session titles, and subsequent tool calls may involve local projects. No local installation or functional test is claimed here.
Why it may be useful
- Nine IM channels and an AI Office Connector in one plugin
- Independent connection, workspace, and session state for multiple bots
- One-command installation for the stable npm release
- Clear README guidance on credential storage and remote-command risks
What to know first
- Each platform still requires separate QR authorization, app creation, or credential setup
- An overly broad bot access scope can expose workspace paths and session metadata
- GitHub source installation may require additional build-script permission
README
dsh-im
Overview
Connect nine IM bot channels to DeepSeek Harness and continue sessions, switch models, manage workspaces, and handle tool approvals from Feishu, WeChat, DingTalk, Slack, Telegram, and other chat clients. Connect IM bots to DeepSeek Harness through QR authorization or bot credentials, with nine channels and an AI Office Connector.
Getting started
- Install the stable npm release into the DeepSeek Harness web profile, restart the Web service, authorize one IM channel, and use the bot's `/status` command to verify the connection.
- dsh-im places nine IM bot channels and the AI Office Connector under one DeepSeek Harness settings entry. Each channel can host multiple bots with separate connection state, workspace, and session bindings.
- Messages can include text and images. The README lists JPEG, PNG, WebP, and GIF sent as an image file, with a 5 MB per-image limit and a 20 MB total image limit per message.
- After restarting `dsh web`, open the plugin settings and confirm that the IM Bots entry is visible. Authorize one channel and wait until the bot is online, then send `/status`. The basic setup is complete when the bot reports that it can connect to DeepSeek Harness.
Configuration
dsh plugin --profile web add -w @xmanrui/dsh-im