GrokBot Desktop Pet Plugin
DeepSeek Harness (DSH) has released a desktop pet plugin that ports the original GrokBot to a Web client, addressing the issue of limited interaction. The plugin supports rich 3D animations, customizable appearances, and physics-based drag interactions, and integrates practical data bubbles that display token usage and tool progress. Users can change outfits, configure settings, and manage notification permissions through the right-click menu. The plugin relies on the official bundled libraries, with its visual data models derived from open-source projects, aiming to provide immersive visual feedback during long conversations.
Read Moreaibo204/dsh-plugin-computer-use: DeepSeek Harness Desktop Interaction Plugin
`aibo204/dsh-plugin-computer-use` is a plugin for DeepSeek Harness (DSH), designed to provide agents with secure, approval-gated desktop control capabilities and audit archiving. By integrating the MCP protocol of Open Computer Use, it supports application management, interactive operations (such as clicking and typing), and high-risk confirmation. Installation requires configuring a Profile Bundle, handling pnpm dependencies, mounting the plugin in an Agent Preset, and setting the appropriate permissions for the target system (macOS, Linux, or Windows). Audit logs can be retained when using `dsh-mcp-client`. The plugin is released under the MIT License, and it should be noted that it only supports Agent Preset.
Read Moredshgirl-starter: Unified DSH Whale Girl Plugin
dshgirl-starter is an unofficial fan-made plugin project based on DeepSeek Harness (DSH), adopting an "everything is a plugin" architecture to unify the Whale Girl persona, desktop interaction features, and memory management. The plugin offers features such as Whale Girl persona injection, a desktop pet, a balance widget, notes, scheduled reminders, right-click to ask DSH, and a shared long-term memory library based on `memory.json`, achieving on-demand memory access via low-token tools. The project supports HTTP API interfaces, facilitating the viewing or creation of notes, reminders, and memories, as well as the initiation of new conversation questions. The runtime environment requires a Node.js version of no less than 22.19 and an x64 CPU architecture; a Python Helper can be optionally used. The source code follows the MIT License.
Read More