PDF and DOCX output - GrantKit Docs

grantkit build compiles your responses into one review document. PDF and DOCX are optional formats behind extras.

Quick start¶

pip install "grantkit[pdf]"      # PDF (WeasyPrint)
pip install "grantkit[docx]"     # DOCX (python-docx)

grantkit build --format pdf --output proposal.pdf  # combined review copy
grantkit build --format docx --output proposal.docx

If the extra isn’t installed, build prints an actionable message and exits with code 2 rather than a stack trace.

How it works¶

build assembles every section (in grant.yaml order) into a single styled review document, then renders it to PDF with WeasyPrint or to DOCX with python-docx. For plain-text portals (accepts_markdown: false) it strips Markdown so the output matches what the portal will accept.

For NSF projects, the combined PDF is visibly labelled as an internal review copy. It uses Arial at 10 points, US Letter portrait pages, one-inch margins, no more than six lines per vertical inch, and no proposer-supplied page numbers. Those formatting defaults do not turn the combined document into an NSF submission artifact. Research.gov requires the applicable proposal sections and attachments to be uploaded or entered separately.

The lower-level grantkit.pdf.PDFGenerator pipeline remains available programmatically for specialized workflows, but callers must still map and validate the resulting files against the applicable Research.gov upload slots.

Page limits¶

grantkit check flags sections whose estimated page count exceeds their page_limit (rule page_limit_estimate_exceeded). The estimate is a rough words-per-page heuristic. An NSF combined review PDF includes review-only material and cannot establish the page count of any separate Research.gov upload; render and validate each upload independently before submitting.

Citations¶

Citations use pandoc-style [@key] syntax. grantkit check reports any [@key] with no matching entry in references.bib (unresolved_citation), but grantkit build does not resolve or format the markers: they remain raw authoring text in its PDF and DOCX review outputs. Prepare submission-ready inline citations and the References Cited upload separately. See Citation management.

Troubleshooting¶

WeasyPrint won’t import. WeasyPrint needs system libraries in addition to the Python package:

# macOS
brew install pango

# Ubuntu
sudo apt-get install libpango-1.0-0 libpangocairo-1.0-0

Fonts. The NSF review stylesheet requests Arial. Make sure Arial is installed wherever WeasyPrint runs; font substitution can change layout and is another reason to treat the combined output as review-only.