DeepSeek-harness-qqbot DSH plugin

This plugin bridges Tencent QQ Bot with DeepSeek Harness, enabling C2C and group message conversations with persistent Harness sessions. It supports image attachments, file sending, Markdown replies, and approval workflows via QQ buttons.

remote-execution

How to install the DeepSeek-harness-qqbot DSH plugin

dsh plugin --profile web add github:sliverp/DeepSeek-harness-qqbot

Copying does not run this command. Review the repository and version before installing the DeepSeek-harness-qqbot DSH plugin.

DeepSeek-harness-qqbot DSH plugin data source

DeepSeek-harness-qqbot DSH plugin snapshot date: Aug 16, 2026

discovered

What the DeepSeek-harness-qqbot DSH plugin can do

Handles C2C and group text messages via official QQ Bot Gateway
Accepts inbound PNG, JPEG, WebP, and GIF images as durable Harness attachments
Automatic text-only fallback when the model does not support images
Outbound assistant text, images, and local workspace files
QQ-scoped `qq_send_file` tool with workspace containment and file-size checks (max 100 MiB)
QQ Markdown replies enabled by default
One persistent Harness session per C2C or group conversation
Agent-preset composition for same tools, prompts, and skills as Web sessions
Proactive QQ approval prompts with Allow/Reject buttons and timeout
Slash commands: /bot-ping, /bot-image-test, /bot-file-test, /bot-help, /bot-status, /bot-cancel
Secrets resolved through Harness credential service
Dormant startup when AppID or AppSecret is not configured

Where the DeepSeek-harness-qqbot DSH plugin fits

Run persistent Harness agents via QQ chat without opening the Web UI
Send images, files, and receive Markdown-formatted responses in QQ
Implement approval workflows for sensitive agent actions through QQ buttons
Create a bot that can handle both text and image inputs with automatic fallback
Manage multiple conversations with separate persistent sessions per user or group

Who the DeepSeek-harness-qqbot DSH plugin is for

Developers building QQ bots with DeepSeek Harness integration
Users who prefer interacting with AI agents through QQ instead of the Web interface
Teams that need to approve agent actions directly in QQ

DeepSeek-harness-qqbot DSH plugin limitations

Requires Node.js 22.19+, pnpm 10.33.4, and DeepSeek Harness 0.1.0-rc.6 or later
Requires a QQ Bot AppID and AppSecret with C2C/group message events enabled; Inline Keyboard permission needed for approval buttons
Only supports official QQ Bot Gateway; no other messaging platforms
File sending limited to 100 MiB per file and only during active QQ message processing
Image input mode 'always' may fail for models that do not accept images
Markdown support requires QQ Bot Markdown permission; otherwise messages are rejected
Approval prompts time out after default 120 seconds
Persistent sessions from Web cannot send files to the previous QQ target

DeepSeek-harness-qqbot DSH plugin: from the repository README

Quoted from the sliverp/DeepSeek-harness-qqbot README, the upstream source of the DeepSeek-harness-qqbot DSH plugin. Copyright remains with the original authors.

An out-of-tree [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) channel plugin that connects an official QQ Bot Gateway to persistent Harness agents. ## Features - Official `@tencent-connect/qqbot-nodejs` Gateway client - C2C and group text messages - Inbound PNG, JPEG, WebP, and GIF images as durable Harness attachments - Automatic text-only fallback when the selected model does not accept image input - Inbound voice transcripts and non-image attachment metadata with the temporary QQ download URL - Outbound assistant text, images, and local workspace files - A QQ-turn-scoped `qq_send_file` tool with workspace containment and file-size checks - QQ Markdown replies, enabled by default - One persistent Harness session per C2C or group conversation - Harness agent-preset composition for the same tools, prompts, and skills as Web sessions - Safe reuse of a live session already opened by Web, without a second session writer - Proactive, requester-bound QQ approval prompts with one-shot Allow and Reject buttons - Approval timeout rejection, fallback to another composed approval channel, and no escalation carry-over to later tool calls - Typing indicators, long-reply

Read the full READMERepository license: MIT

DeepSeek-harness-qqbot DSH plugin questions

How do I install the DeepSeek Harness QQBot plugin?

Run `pnpm dsh plugin --profile web add github:sliverp/DeepSeek-harness-qqbot` in your DSH environment. Make sure you have Node.js 22.19+, pnpm 10.33.4, and DeepSeek Harness 0.1.0-rc.6 or later installed.

Can I use this plugin without configuring AppID and AppSecret?

Yes. The plugin starts in a dormant state if AppID or AppSecret is missing. It will not block DSH from completing its startup. Once you configure both values and reload or restart DSH, the QQ channel will connect.

Does the plugin support sending files from the agent to QQ?

Yes. The agent can use the `qq_send_file` tool to send files up to 100 MiB. The file must reside within the agent's working directory. The tool is only active while processing the current QQ message, so it cannot be used from a Web session targeting the same QQ user.

How does the approval workflow work in QQ?

When an agent action requires approval, the plugin sends a proactive QQ message with 'Allow once' and 'Reject' buttons. Only the QQ user who initiated the turn can decide. The grant covers only that exact operation. If the prompt times out (default 120 seconds), the approval is rejected. If sending the prompt fails, the plugin falls back to another composed approval channel (e.g., Web).

What image formats are supported?

The plugin supports inbound PNG, JPEG, WebP, and GIF images as durable Harness attachments. If the selected model does not accept image input, the plugin automatically falls back to text-only mode and includes a metadata notice with the temporary download URL.