dsh-dialogue-mindmap
維護狀態: 活躍gaor0605-kq/dsh-dialogue-mindmap
DSH plugin: a mind-map canvas beside the Chat tab — branch follow-ups from any turn in the same session. / 在「对话」旁加画布视图,从任意节点在同一会话里继续拓展。
$ dsh plugin add dsh-dialogue-mindmap0
星數
0
Fork
JavaScript
語言
MIT
授權條款
2026-11-01
建立於
2026-11-01
最近推送
README
dsh-dialogue-mindmap
English | 中文
A DeepSeek Harness (dsh) Web UI plugin that adds a Canvas view beside the Chat tab of a conversation. Every turn of the session is drawn as a node in a mind map, and you can branch a follow-up from any node — inside the same session, without creating a new one.
Features
- Canvas tab next to the Chat tab; click to switch between the two views of the same conversation. While the Canvas is shown, the composer below is hidden and the canvas fills the remaining height.
- One node per turn, laid out as a left-to-right tree. The first turn is the root; by default each new turn hangs under the previous one.
- Branch from any node: select a node, type a follow-up in the side panel and press the button (or Ctrl+Enter). The message is sent to the same session, prefixed with a short reference to the selected turn, and the resulting turn is attached as a child of that node. A dashed "waiting for answer" placeholder shows until the turn arrives.
- Detail panel: clicking a node opens a side panel with the full question and the full answer (scrollable), a button to switch back to the Chat view, and the branch input. Drag the panel's left edge to resize it (double-click the edge to reset); the width is remembered.
- Pan and zoom: drag empty canvas space to pan; mouse wheel zooms around the cursor (25%–250%); the bottom-right controls zoom out / reset / zoom in.
Install
dsh plugin add dsh-dialogue-mindmap
Then refresh the Web UI and open any conversation — a Canvas tab appears next to Chat.
Limitations
- The tree structure (which turn is a child of which node) is stored in the browser's
localStorage, per session. It is not synced across browsers or devices; clearing site data resets the layout to a linear chain. - Because branching happens inside one session, the model still sees the whole conversation history; the prefix only tells it which earlier turn the follow-up builds on.
- Full question/answer text in the detail panel is read from turns already loaded in the Chat view. For very old turns that are not loaded yet, the panel falls back to the truncated preview — scroll up in the Chat view once to load them.
- Hiding the composer relies on the Web UI's
data-composer-seatattribute and may need an update if the host UI changes.
Requirements
DeepSeek Harness Web UI (tested on dsh 0.2.1-rc.5). No host-side dependencies; the plugin is client-only.
License
更多「知識與研究」外掛
weknora
作者 tencent
开源的 LLM 知识库平台:把原始文档转化为可检索的 RAG 知识库、自主推理智能体和自动维护的 Wiki 系统,支持多租户部署。
honcho
作者 plastic-labs
用于构建有状态 AI 智能体的记忆库,支持跨会话连续学习与上下文工程。
webnovel-writer
作者 lingfengqaq
基于 Claude Code 的长篇网文辅助创作系统,解决 AI 写作中的「遗忘」和「幻觉」问题,支持 200 万字量级 连载创作。
billion-context
作者 ranxianglei
A context-compression plugin for month-long single sessions (billions of tokens) ,small context windows (a 100K context is enough) and token savings (5x fewer tokens) .上下文压缩插件,月级长会话会话, 兼顾小窗口(100k上下文足矣)省token(省5倍token) . billion-context is all you need
