Skip to content

docs: task-first Jupyter Book site (OLED, agents, CI) - #61

Open
bobleesj wants to merge 28 commits into
mainfrom
jupyter-book-docs
Open

bobleesj wants to merge 28 commits into
mainfrom
jupyter-book-docs

Conversation

@bobleesj

Copy link
Copy Markdown
Owner

Summary

  • Restructure docs into three science-facing tutorials: physical features from a .cif, statistics over many CIFs, and the Oliynyk OLED table (full list + property meanings).
  • Add agent/human surface: llms.txt, API quick-reference, CITATION.cff / static citation paths next to data access; widen layout (+200px), drop right Contents.
  • CI: docs workflow builds + e2e check and deploys to gh-pages; remove Codecov from matrix/PR workflows and drop .codecov.yml.
  • Small API helpers on Oliynyk (to_dataframe / to_csv) and docstrings so the OLED tutorial stays copy-paste friendly.

Test plan

  • Local jupyter-book build docs + scripts/docs_e2e_check.py PASS
  • CI docs workflow green on this PR
  • After merge: live site serves new tutorials, _static/oled.csv, llms.txt
  • Spot-check that docs/_build/ and duplicate docs/data/oled.csv were not committed

bobleesj and others added 4 commits July 10, 2026 18:16
Restructure tutorials around physical features, multi-CIF stats, and
the Oliynyk OLED table; widen layout; add citation/llms surface; drop
Codecov from docs UX; publish via Actions e2e + gh-pages.
Plug at top of intro: cifkit is built with scikit-package for
maintainable, agent-friendly scientific packaging; add soft cite row,
BibTeX, CITATION.cff, README, and llms.txt.
Matches requires-python / classifiers already on main after #58.
@bobleesj
bobleesj force-pushed the jupyter-book-docs branch from 406f6cc to 9f31d62 Compare July 11, 2026 01:17
bobleesj added 24 commits July 10, 2026 18:18
Add Selvaratnam et al., Data in Brief 63, 112138 (DOI
10.1016/j.dib.2025.112138) with ScienceDirect PII S2352340925008595
to CITATION.cff, CITATION.txt, intro, OLED tutorial, README, llms.txt.
Expand physical-features with step-by-step CN determination and best-
method selection; embed a Three.js widget with real GdSb coordinates;
keep demo/cite boilerplate at page bottoms and on the home page; stress
that OLED is the Data in Brief table, not values from a .cif.
Lead with the research need: reliable Python supercells, interatomic
geometry, CN shells, and structural featurization over thousands of
CIFs — simple Cif / CifEnsemble API for experimental and ML workflows.
PyPI 1.2.2 already classifies 3.12–3.14; img.shields.io/pypi/pyversions
was still serving a 6h-cached 3.11–3.13 SVG. Use an explicit badge.
Rewrite benefit framing as scientist-facing goals (help extract
features, reliable supercells, high-throughput folders), softer cite
language, and consistent scikit-package acknowledgement wording.
Replace U+2014/U+2013 punctuation with ASCII hyphen spacing for a
consistent, scikit-package-like plain-text voice across docs and README.
In How-cifkit-benefits, note published high-throughput featurization
via SAF/CAF and the Digital Discovery paper DOI.
Add Digital Discovery paper (10.1039/d4dd00332b) to Publications,
README, and CITATION files; note cifkit as SAF geometry engine.
Frame as folder Q&A (what is in here, filter, copy, histogram), clarify
demo (GdSb/HoSb) vs published large-ensemble figures, and link SAF/CAF scale.
Single Maintainers entry: news files, cutting a release, docs deploy,
and CHANGELOG.rst include. Drop the separate changelog.md toc entry.
Nest API pages under api/index and release under maintainer/index
(quantem.widget chevron pattern; no caption on those parts).
Note tested db_source detection and preprocess for ICSD/COD/PCD/MP/CCDC/MS
in the scientist-benefits pitch, README, and physical-features tutorial.
Captioned parts again so API reference and Maintainers are not peer
links under Tutorials; keep expandable Overview children under each.
Captioned sections list API pages and Changelog/release directly so
sidebar is not Tutorials + fake Overview parents.
Reconstruct URhIn (ZrNiAl) from test expectations, ship open multi-source
stubs (ICSD/COD/MS/MP/CCDC/PCD), seed folder fixtures, update atom-count
expectations, and convert missing-fixture FileNotFoundError to skips so
pytest is green without the full private PCD dump.
Force off-screen Plotter, screenshot then close (no bare show()); set
headless:true on PR/matrix workflows so Xvfb is available if needed.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant