PhysicalAI-0/dsh-paperforge 预览 preview

PhysicalAI-0/dsh-paperforge

PaperForge — forge your understanding of papers inside DSH: selection-to-highlight, selection-to-ask (one dedicated session per paper), marquee-to-note screenshots with backlinks.

Project Overview项目介绍

PaperForge is a DSH-native plugin purpose-built for reading, annotating, and taking notes on academic PDFs inside DeepSeek Harness. It installs with a single command, dsh plugin --profile web add github:PhysicalAI-0/dsh-paperforge, and requires DSH version 0.1.7 or newer together with pnpm available on the host PATH. Once installed, launching dsh web inside any directory that already serves as a DSH workspace adds a "论文" entry to the left sidebar and unlocks a native three-column reading layout, with the workspace's PDF list and per-paper chat threads on the left, a dedicated AI conversation for the opened paper in the middle, and the PDF plus a Markdown notepad arranged side by side on the right.

The intended workflow starts by clicking any paper to enter the reading view: users select a passage to trigger "Ask AI", which prepends an @论文 reference plus the current page number and selected text into the conversation, or highlight spans in six colors and drag a rectangle over figures to drop a screenshot directly into the notes. Highlights, screenshots, and chat bindings are persisted as highlights.json, session.json, and image files under paperforge_notes/<论文名>/, while any .md block can be clicked to edit source inline, autosaving on blur and merging changes made by the plugin or the AI. The Summary tab batches selected sessions three at a time and lets the model write a consolidated Markdown digest, and the experience is aimed at researchers who want a local, replayable reading history.

Dependencies are minimal: DSH 0.1.7+, pnpm for installation, and an optional Python sidecar (py/) only when highlights must be written back into the PDF file itself. Known limits include a maximum of two right-hand columns, undo scoped to the current block, broken cross-block reference links and footnotes, .md files outside the workspace or larger than 2 MB becoming preview-only, and a DSH cap of three cited sessions per message that the plugin works around by chunking. The plugin ships under the MIT license, but configuration should not include - insert: or it will load twice, and any model specified in cordis.patch.yml becomes DSH's default for new chats.

PaperForge 是为 DSH(DeepSeek Harness)专门打造的原生论文阅读插件,通过 dsh plugin --profile web add github:PhysicalAI-0/dsh-paperforge 一行命令即可安装,并要求 DSH 版本 ≥0.1.7 与 PATH 中可用 pnpm。安装后在工作区左侧栏出现"论文"入口,呈现左中右三栏布局:左边为工作区 PDF 列表与每个论文的会话列表,中间为该篇专属 AI 会话,右栏并排显示 PDF 预览与 Markdown 笔记,深度复用 DSH 原生的预览、对话与模型配置能力。

典型流程是在 PDF 目录中启动 dsh web,点开任意论文进入阅读视图,用六色划词高亮与框选截图直接落入 paperforge_notes/<论文名>/ 目录下的 Markdown 笔记;选中段落可触发"Ask AI",把 @论文、页码与选段上下文注入当前会话,便于追问与回看。Summary 功能会按每三条消息一批勾选式生成综合 Markdown;点开任意 .md 即可按块就地编辑源码,离开即自动保存与渲染。它面向需要在本地持续阅读、做批注与积累笔记的科研读者与从业者,灵感来自 alphaXiv 的划词即问交互,但摆脱其额度与模型限制。

依赖方面除 DSH 0.1.7+ 与 pnpm 外,可选 Python 边车用于把高亮写回 PDF;限制包括右侧栏最多两栏、跨块撤销与引用式链接受限、超过 2 MB 或工作区外的 .md 仅可预览、Summary 受 DSH 每消息最多引用 3 个会话约束,许可证为 MIT。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 1 warning1 项注意
  • Only 2 stars - very few users, little community feedback星标只有 2,几乎没人在用,遇到问题缺少社区反馈
DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

  • DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
  • External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时(Electron / Python / Docker 等)是否齐备
  • Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突:命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

  • Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致,是否归档或长期停更
  • Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
  • curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
  • Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
  • Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需(filesystem / network / shell / clipboard)
  • Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限,以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照,不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装(适合开发者)

CLI Install命令行安装

dsh plugin --profile web add github:PhysicalAI-0/dsh-paperforge

把 PhysicalAI-0/dsh-paperforge 加入你的 DSH 配置(web profile)即可启用。

READMEREADME

PaperForge:水彩风格的论文、对话气泡与钢笔

在 DSH 里读论文:边读、边问、边记

快速开始 · 亮点 · TODO · 开发手册 · English

License: MIT DSH Plugin DSH Repo

读懂一篇论文,留下自己的思考。 从论文库点开 PDF,原文、专属 AI 会话与 Markdown 笔记并排展开:选中难懂的段落,带着上下文向 AI 追问;划下重要的文字,框选值得留存的图表,高亮与截图自动收进本地笔记。下次打开,接着上次的思考往下读。

News

  • 2026-09-29
    • 原生多栏阅读布局:左边会话列表,中间本篇论文的 AI 会话,右边 PDF|笔记。
    • 新增会话 Summary 与复制路径;笔记目录改为 paperforge_notes,不再引用的图片自动清理。
    • Markdown 块级编辑:所有 .md 都能原地编辑、自动保存。
    • 阅读链路重做:论文库、笔记实时预览、每篇论文一个专属 AI 会话。
  • 2026-09-28 — 首个版本发布;改为 DSH 原生 dsh plugin 安装。

亮点

阅读布局:左边会话列表,中间本篇论文的 AI 会话,右边 PDF 与笔记

论文库
左侧栏「论文」列出工作区里的 PDF,点一篇进入阅读布局
标注
划词高亮(六色)与框选截图自动写进笔记;选段也可复制为原文或 Markdown 引用
Ask AI
划词「Ask AI」或页头「AI」:问题带上 @论文、页码与选段,发进本篇论文的会话
Markdown 编辑
所有 .md 都能原地编辑:点一块改源码,离开即渲染;自动保存,与插件、AI 的改动自动合并
Summary
论文库「会话」页签勾选会话,生成一份 Summary(Markdown):每 3 个一批引用,最后综合
复制路径
任意文档页头的复制图标,一键复制文件路径
点一块即可编辑源码,其余保持渲染
笔记:点一块改源码,其余保持渲染
Summary:AI 把多个会话的总结写进 Markdown
Summary:勾选的会话由 AI 汇总成一份 Markdown

Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要,完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →

← 上一个 Prev dsh-memory-amem 下一个 Next dsh-tool-visual-primitives →