Generate Chinese flashcards from raw learning text using Wiktionary and OpenAI.

• Python 3.9+
• OpenAI API key

1. Install dependencies:

python3 -m pip install -r requirements.txt

2. Create a .env in project root:

OPENAI_API_KEY=sk-...
# Optional model override
OPENAI_MODEL=gpt-4o

• Create instance directories under output/, each with an input.txt:

mkdir -p output/book
cp your_input.txt output/book/input.txt

• Run generator (streams progress, skips existing .md files, halts on first error):

.venv/bin/python generate.py --verbose

• Setup venv + deps: make setup
• Run generator via Makefile:

make generate

• Per-word files: <HEADWORD>.md in each instance directory under output/.

• Automatically creates/uses extracted.txt in each instance to list vocab; edit it to change the processing set/order.
• Skips any vocab that already has <HEADWORD>.md in the instance directory.
• Recurses for multi‑character words: writes parent, subword character cards, and component cards until no more named components.
• Halts on the first BLOCKED with a detailed reason; re‑run to resume (completed files are skipped).
• Pronunciation is Pinyin‑only (tone marks), multiple readings separated by /.
• Examples are formatted as ZH (pinyin) - EN.

• .env is auto-loaded on startup (no manual export needed).
• The tool calls OpenAI for judgment steps (headword extraction and single-headword field extraction), then validates and writes Markdown following a fixed schema.