Paper Reader

A local web app to read PDF papers side-by-side with an automatic Chinese (or other-language) translation, powered by pdf2zh.

Features

• Side-by-side panes: original PDF on the left, translated PDF on the right, with synced scrolling.
• Clickable links: internal page jumps and external URLs both work in the rendered view.
• Auto-numbered table of contents: outline panel generated from the PDF's bookmark tree.
• Selectable text layer: copy text from both the original and translated panes.
• Bounded directory browser: serves only files under the configured directory; no path-traversal escapes.

Requirements

• Python >= 3.11

pdf2zh is installed automatically as a dependency.

Install

pipx install paper-reader        # once published
# or from a local checkout:
pipx install .

After installing, run paper-reader doctor once to apply two colorspace bug-fix patches to the bundled pdf2zh/pdfminer (without them, color figures may render red). This is idempotent — safe to run repeatedly.

paper-reader doctor

Why is this needed? pdf2zh 1.9.11 (the current release) still ships both colorspace bugs. The doctor command patches them in-place so you do not have to wait for an upstream fix.

Usage

paper-reader [DIR] [options]

• With no DIR, Paper Reader serves the current working directory.
• paper-reader ~/papers serves that directory.
• Open the URL printed on startup (default: http://127.0.0.1:8733).
• Pass --open to have the browser open automatically.

Examples:

paper-reader                      # serve current directory
paper-reader ~/papers             # serve ~/papers
paper-reader ~/papers --open      # serve ~/papers and open browser
paper-reader --port 9000          # use a different port

Configuration

Settings are resolved in this order (highest priority first):

CLI flag > environment variable > config file > default

The config file location is ~/.config/paper-reader/config.toml (respects XDG_CONFIG_HOME).

Example ~/.config/paper-reader/config.toml

port = 8733
host = "127.0.0.1"
engine = "google"
lang_from = "en"
lang_to = "zh"
concurrency = 4
cache_dir = "~/.cache/paper-reader"

How It Works

Flask serves a pdf.js frontend that renders two PDF panes side-by-side. Translation is performed per-page by invoking the external pdf2zh CLI as a subprocess; translated PDFs are cached so each page is only translated once.

License & Credits

This project is licensed under the MIT License (see LICENSE).

It invokes pdf2zh / PDFMathTranslate (AGPL-3.0) as a separate command-line program via subprocess. This is aggregation, not a derivative work — the Paper Reader source code remains MIT while pdf2zh's AGPL obligations remain with pdf2zh itself.