跳到正文
GitHub·· 9 天前AI 评分38

EasyPaper:用 AI 翻译、阅读、标注和讨论 PDF 的跨平台工作区

orion-gz/EasyPaper

AI 导读

EasyPaper 是一个用于 AI 翻译、阅读、标注和讨论 PDF 的 Web 与桌面工作区,支持本地 Ollama、Gemini、Claude、OpenAI 以及 Antigravity、Claude Code、Codex 等 CLI 提供商。

正文

EasyPaper is a web and desktop workspace for translating, reading, annotating, and discussing PDFs with AI. Use local Ollama, Gemini, Claude, OpenAI, or supported CLI providers including Antigravity, Claude Code, and Codex.

EasyPaper research mode dashboard Research mode workspace

Two reading modes

Mode For What it emphasizes
Research mode Papers, surveys, theses, preprints, and academic reports Paper structure, methods, evidence, citations, research graphs, comparisons, and pre-reading briefs
General document mode Technical documentation, books, articles, reports, manuals, policies, presentations, and other general documents Document-aware translation, summaries, vocabulary, focused Q&A, full-text search, and document overviews

Both modes currently accept PDF files. General document mode lets you use features appropriate to the purpose and type of each document.

More screenshots

Research mode dashboard General document mode dashboard
Research mode dashboard General document mode dashboard
Viewer with AI assistant Viewer image overlay
Viewer and AI assistant Figure and image reference overlay
Viewer memo Viewer reference overlay
Annotations and floating memos Citation reference overlay
Document library AI Chats
Library AI Chats
Research graph heatmap Reading records
Research graph heatmap Reading records

Features

  • Side-by-side source and AI translation with sentence-level alignment
  • Context-aware AI chat, pre-reading briefs, summaries, and vocabulary support
  • Research workspace with library, AI Chats, research graph, reading analytics, and recommendations
  • Annotations, highlights, floating memos, and annotated PDF export
  • Reference, figure, table, and equation overlays for research PDFs
  • Native desktop apps for Windows, macOS, and Linux, plus a self-hosted web app

Quick start

Desktop app

You can download the installer for your platform from the latest release. The Tauri desktop app includes the backend sidecar, so no additional Python or Node.js installation is required.

OS Installer
Windows .msi or .exe
macOS (Apple Silicon) _aarch64.dmg
Linux .AppImage, .deb, or .rpm

At first launch, choose and configure an AI Provider in the onboarding flow. Unsigned desktop builds can trigger an operating-system security warning.

Important

Note macOS: Because the app is not fully signed, recent versions of macOS (Ventura and later) may show the misleading message “The application cannot be opened because it is damaged” instead of allowing you to bypass the warning with right-click → Open. The file is not actually damaged; this is caused by the quarantine attribute added during download. Remove it in Terminal with the command below, then try again.

xattr -cr /Applications/EasyPaper.app

Windows: On the “Windows protected your PC” screen, select More info → Run anyway to install normally.

Run from source

Requirements: Python 3.8+, Node.js 16+, and npm. Ollama is optional. The scripts below make setup and startup straightforward.

git clone https://github.com/orion-gz/EasyPaper.git
cd EasyPaper
./scripts/sh/setup.sh
./scripts/sh/start.sh

On Windows, run scripts\bat\setup.bat, then scripts\bat\start.bat. Open http://localhost:8000 after the server starts.

Docker

git clone https://github.com/orion-gz/EasyPaper.git
cd EasyPaper
docker compose up -d --build

Open http://localhost:8000. Data is persisted in the /data volume. To stop the service, run docker compose down.

Initial account

Field Value
Username admin
Password admin

Change these credentials in Settings after signing in.

Testing

cd backend
.venv/bin/pip install -r requirements-dev.txt
.venv/bin/python -m pytest tests/ -v
cd frontend
npm install
npm run build
npm run test:e2e

CLI-based AI Providers

EasyPaper detects installed and authenticated agy, claude, and codex CLIs at startup. Select any detected provider from the model picker; API providers and Ollama remain available as alternatives.

See CHANGELOG.md for release notes.

来源:GitHub · github.com