Translate

Use interactive setup when a TTY is available:

node scripts/init-config.js --interactive

Use flags when you already know the preferences:

node scripts/init-config.js \
  --default-target-language zh-CN \
  --default-mode faithful \
  --tone neutral \
  --preserve-formatting true

Workflow

1. Detect the source language and the requested target language.
2. If the target language is not specified, translate to zh-CN.
3. Resolve preferences from the current task and the user config.
4. Pick the translation mode:
   • literal: quick, close to the source wording
   • faithful: accurate and natural, default
   • localized: adapt tone and phrasing for the target audience without changing facts
5. Preserve structure:
   • keep headings, lists, links, tables, and code fences intact
   • do not translate code, commands, file paths, or identifiers unless the user asks
   • apply glossary and do_not_translate consistently
6. For long inputs, split them before translating:

node scripts/chunk-text.js --input <file-or-> --max-chars 2200

Translate chunk by chunk, then merge:

node scripts/merge-output.js --input <translated-chunks.json>

Long Text Handling

Use chunking when the input is long enough that consistency or context window pressure becomes a risk.

• Prefer chunk boundaries at paragraph breaks.
• Keep each chunk structurally coherent.
• After chunked translation, scan the full merged output for:
  • inconsistent terminology
  • broken list numbering
  • duplicated headings
  • damaged Markdown or code fences

Reference

Read these files as needed:

• references/modes.md for mode selection and output expectations
• references/config.md for config schema and CLI usage
• references/style-guide.md for formatting, glossary, and technical-text rules