Get started¶
Paper Espresso is currently an early, Codex-first skills-only plugin. The canonical workflow is skills/paper-espresso/SKILL.md.
Start a processing pass¶
A paper title, arXiv ID, arXiv URL, or local PDF is the only input required to start. Choose exactly one of the following alternative launch lanes. They start the same workflow; do not run both.
Installed plugin¶
Use this lane for normal operation:
Repository checkout¶
Use this lane only while developing from the repository before installing the plugin:
In either prompt, replace PAPER with a paper title, arXiv ID, arXiv URL, or local PDF path.
With no other instructions, Paper Espresso uses these defaults:
- at most one US Letter page (request A4, another size, or an exact count when needed);
- emphasis inferred from the paper;
- a technically literate reader outside the paper's immediate subfield;
- editable
digest.texand a compiled, layout-checkeddigest.pdf.
Everything after PAPER is optional and overrides a default. For example:
If a title matches multiple papers, Paper Espresso asks for clarification before continuing.
Requirements¶
- Python 3.10+
- uv
- Poppler:
pdftotext,pdfinfo, andpdftoppm - Tectonic,
pdflatex,lualatex, orxelatex
The runtime itself uses Python's standard library.
For local marketplace installation during development, consult the official Codex plugin guidance.
Expected output¶
digest.tex: editable LaTeX with no dependency on temporary downloadsdigest.pdf: compiled and layout-checked within the requested maximum page count
Downloaded source and intermediate files live in a marked temporary workspace and are deleted after completion.
Build this documentation¶
For a production build:
The generated site is written to site/.
Publish with GitHub Pages¶
The included workflow deploys the documentation after relevant pushes to main. In the GitHub repository, configure Settings → Pages → Build and deployment → Source to GitHub Actions once. The workflow then builds and deploys the site automatically.