dsh-qqbot

No matter how capable a DeepSeek Harness agent is, the friction of reaching for a terminal or WebUI before every interaction adds up. dsh-qqbot removes that step by turning QQ into a first-class frontend for your dsh agent: send a message in a private or group chat from your phone, and the agent handles tool execution, model switching, and session management in the background, then streams the reply back through QQ.

Beyond chat, the plugin addresses the details that matter in daily use. Group chats support @bot triggering, quoted-message context, and file-attachment downloads; private chats deliver streaming replies with Markdown chunking. Sessions are isolated per peer, reclaimed after idle timeouts, and restored across restarts. Model routing is flexible—set a per-peer preference, configure via settings.yaml, or fall back to the host default. Built-in commands such as /bot-model, /bot-reset, /bot-status, and /goal cover the most common operations. The approval bridge keeps workflows tight: when the agent needs confirmation, a button or text prompt appears in QQ while the WebUI shows it in sync, so you never have to hop between surfaces. Long-running goal tasks send a completion notification when they finish.

If you live in QQ, want to invoke your dsh agent from a phone or tablet, or prefer handling approvals inside an IM chat rather than a terminal, dsh-qqbot is ready to go out of the box. It is written from scratch with no reliance on Tencent-internal code; the underlying protocol uses the MIT-licensed @tencent-connect/qqbot-nodejs, making it compliant, self-hostable, and easy to audit.