langreader

langreader

Generate language-learning HTML readers from Markdown using an LLM.

langreader is a CLI tool that converts Markdown texts into interactive HTML readers. It uses Large Language Models (LLMs) via multiai to split paragraphs into natural sentences and provide sentence-level translations.

👀 Demo

Before reading further, check out what this tool generates:

👉 Live Demo: French Reader (Alsace)

Features

  • 🤖 AI-Powered Translation: Automatically splits paragraphs into natural sentences and translates them line-by-line.
  • 🔊 Text-to-Speech: Built-in audio buttons for every source sentence (uses browser Web Speech API).
  • ⏯️ Resumable Generation: Hit an API rate limit? Use --continue to resume generation exactly where it stopped.
  • 📝 Markdown Support: Preserves headings, code blocks, and basic formatting.
  • 🌍 Wikipedia Helper: Includes a script to easily fetch and format Wikipedia articles for learning.

Installation

pip install md-llm-lang-reader

Prerequisite: multiai configuration

This tool uses multiai internally to access LLM providers (OpenAI, Anthropic, Google, etc.). Before using langreader, you must configure multiai.

Details on supported providers and configuration can be found here: https://sekika.github.io/multiai/

Quick Start

  1. Prepare a Markdown file (input.md): 
    # Bonjour

Ceci est un exemple de phrase.
  2. Run the command: 
    langreader -i input.md -o output.html --src fr --tgt en --provider openai --model gpt-4o-mini
  3. Open output.html in your browser.

Documentation Navigation

  • Usage Guide: Detailed CLI options, YAML metadata, and styling.
  • Tutorial: A step-by-step example creating a French reader from Wikipedia (“Alsace”).
  • TTS Setup: How to configure voices on Windows, macOS, iOS, and Android.