🌿 pdfgen v1.1.0 · MIT

Markdown + template → PDF, with an automatic post-generation verifier: no silent tofu, no missing content, ever.

What it does

Verified output: every PDF is re-opened after generation and its text extracted (PyMuPDF/fitz). 0 U+FFFD "tofu" characters, 100% of expected content lines present (whitespace-normalized matching), embedded fonts reported.
CJK-first: templates ship CJK font stacks (Songti SC) — Chinese renders correctly out of the box.
3 templates: report, invoice, product_sheet; bilingual FR/中文 fixtures included.
JSON verification report: pages, fonts, tofu hits, missing lines, CJK presence.
Self-testing: includes a negative test proving the verifier really detects anomalies. Exit codes: 0 verified OK · 1 verification failed · 2 bad input.

Install (pip, v1.1+)

pip install git+https://mandrilly.com/git/pdfgen.git

Installs a pdfgen command (deps: weasyprint, markdown, PyMuPDF; on macOS export DYLD_LIBRARY_PATH=/opt/homebrew/lib so pango is found):

pdfgen --template product_sheet --meta meta.json --body body.md -o out.pdf

Quick start (from source)

python3.12 pdfgen.py \
    --template report \
    --meta fixtures/report_meta.json \
    --body  fixtures/report_body.md \
    -o /tmp/report.pdf --verify-report /tmp/report.verify.json

Requires Python 3.12 with weasyprint, markdown, PyMuPDF (fitz); pango installed (brew install pango on macOS).

Tests

bash run_tests.sh   # 3 templates + negative test → "TOUS LES TESTS OK"

Source

git clone https://mandrilly.com/git/pdfgen.git