David's Filetranslator Skill

A skill for translating files with a terminology-first workflow and format-preserving output.

Note: the current document parsing is powered by basic Python scripts and still needs optimization, so use with care.

Features

• Supported formats: docx, xlsx, pptx, pdf
• Translation pipeline:
  1. Parse source file into segments
  2. Extract key terms
  3. Translate glossary first
  4. Translate all segments with glossary lock
  5. Write back translated output
  6. Run QA checks
• OpenAI-compatible API provider support

Folder Structure

• SKILL.md: skill metadata and instructions
• agents/openai.yaml: UI metadata
• scripts/: pipeline scripts
• references/: glossary and format policies

Requirements

• Python 3.9+
• pypdf
• python-docx

Install dependencies:

python3 -m pip install -U pypdf python-docx

Quick Start

cd /Users/aroma/Desktop/Openclaw/skills/davids-filetranslator-skill

export OPENAI_API_KEY="YOUR_API_KEY"
export OPENAI_BASE_URL="https://generativelanguage.googleapis.com/v1beta/openai"
export FILETRANSLATOR_OPENAI_MODEL="gemma-3-12b-it"

python3 scripts/run_pipeline.py \
  --input /path/to/file.pdf \
  --source-lang en \
  --target-lang zh \
  --domain "general" \
  --provider openai \
  --output-dir /tmp/david-filetranslator-output

Notes

• For scanned PDFs, run OCR first.
• Current PDF writeback keeps original PDF and outputs translated sidecar text.

License

MIT