Edwardxlai/easyread

What it solves

EasyRead is designed to make reading English academic papers more comfortable for Chinese speakers. It solves the friction of using generic translation software by providing a formatted, side-by-side reading experience that preserves formulas, tables, and original layouts while allowing for deep interaction with AI and personal note-taking.

How it works

Users import PDFs (via file upload, arXiv ID, DOI, or URL). The system translates the document page-by-page in the background using a chosen LLM engine. It renders the translation in a clean, book-like format using KaTeX for formulas and three-line tables for data. The application runs locally, ensuring that papers and notes remain on the user's machine.

Who it’s for

Researchers, students, and academics who read large volumes of English scientific literature and want a streamlined workflow for translation, annotation, and AI-assisted comprehension.

Highlights

• Integrated AI Interaction: A dedicated "Ask AI" panel allows for real-time dialogue, where users can drag selected text into the input box to ask specific questions about the paper.
• Dual-View Reading: Supports a "Comparison" mode (English and Chinese side-by-side) and a synchronized original page view that highlights the current paragraph's location.
• Flexible Model Support: Compatible with a wide range of engines including Claude Code, Codex CLI, various cloud APIs (DeepSeek, OpenAI, Gemini, etc.), and local models via Ollama or LM Studio.
• Detailed Annotation System: Features a four-color highlighter, underlining, and the ability to export notes as Markdown for tools like Obsidian or Notion.
• Local-First Privacy: All data is stored locally in JSON files, avoiding the need for a central database and allowing for easy backups.
• Agent Integration: Includes a command-line interface that allows AI agents (like Claude Code) to read notes and write responses directly into the paper's margins.