OpenADMET website

This repository contains the source code for the OpenADMET website, which is built using Hugo and is deployed to GitHub Pages. We are using Cloudflare Pages to handle preview deployments of changes.

Preview the website locally

0. Download and install Hugo.
1. In the top-level directory, run hugo server -D
2. Copy the Web address from this line (in this example, it is localhost:1313):

Web Server is available at //localhost:1313/ (bind address 127.0.0.1)

3. Paste the Web address into your browser

Making changes

1. Create a new branch with git checkout -b <mybranch>, where <mybranch> is the name of your new branch.
2. Commit the changes (git commit .) and push the repository to your branch (git push origin <mybranch>)
3. Open a PR and request review

Deploying

GitHub Pages (Current)

Deploying is handled by a Github actions (see the .github/workflows/gh-pages.yml file). Commits/PRs into the main branch will trigger a new build. Once built, the action will push the content to the gh-pages branch, which the website is now served from.

Cloudflare Pages (Optional)

The repository is also configured for Cloudflare Pages deployment with automatic PR previews. This provides:

• PR Previews: Every pull request gets a unique preview URL for testing changes
• Cleanup: Automatic removal of old preview deployments when PRs are closed

Cloudflare Workflows

• build-pr.yaml - Builds Hugo site for PR previews
• stage-cloudflare.yaml - Deploys PR previews to Cloudflare staging
• cleanup-cloudflare.yaml - Cleans up old preview deployments

Developer Guide

Adding new members

1. Add an image to content/members/img/ (e.g. firstname-lastname.jpg).
2. Add a YAML file firstname-lastname.yaml to data/members/ in the following format:

name: John D. Chodera
role: Primary Investigator
lab: Chodera lab
title: Associate Member
institution: Sloan Kettering Institute
img: john-chodera.jpg
webpage: "http://choderalab.org"
description: Automated Bayesian force field fitting and prediction of systematic error
google_scholar: http://goo.gl/qO0JW
ORCID: 0000-0003-0542-119X
twitter: jchodera
github: jchodera
active: true

You can remove someone from appearing by setting active to false

Adding a job

1. Add a YAML file to /data/jobs with the following format

role: Role Title
title: Title
anchor: title-blah
institution: Placeholder Institution
location: Remote
investigator: Mary Appleseed
investigator_url: https://investigator.example
contact_email: email@example

availability: Immediate
status: open
summary: |
  Job description

To set the job to unavailable, modify the status tag to closed

Auto-updating Blog & Videos cards

The homepage Blog card, the homepage Videos card, and the full /videos gallery are not hand-edited. They are generated at build time from live feeds, and the nightly build (see .github/workflows/build-push.yaml) refreshes them automatically — so a new blog post or YouTube video appears on the site within a day of publishing, with no code change.

How to control what a card says

• Blog card text comes from the post's Excerpt field in the Ghost editor. If you leave the Excerpt blank, Ghost falls back to scraping the opening sentences of the post, which usually reads poorly on a card. Set the Excerpt when you publish and that becomes the card copy.
  • External-link posts are a special case. A few Ghost posts are stubs that redirect to a self-hosted article elsewhere (e.g. "Navigating PXR Chemical Space…" redirects to openadmet.github.io/octant-pxr-htchem-blogpost/). These have no body in the RSS feed, so there is nothing for Ghost to scrape — if such a post is the newest one, the card preview goes blank unless you set an Excerpt. Always set the Excerpt on external-link posts. The redirect itself is unaffected: the card links to the Ghost URL, and Ghost forwards to the self-hosted page, so the "More →" button keeps working as before.
• Video card text comes from the video's description on YouTube. That description is the single place to edit a video card. A video with no description falls back to generic text, and the build prints a WARN naming the video so you know to add one.

Where the feed URLs live

All three feeds are configured under [params] in config.toml (ghostFeed, youtubeFeed, youtubeChannel). Point youtubeFeed at a different channel by changing its channel_id.

Known behaviors

• Dates render in UTC. A video uploaded late in the Pacific evening will show the next day's date. This is not auto-corrected: Hugo ignores the site timezone once a feed timestamp carries a UTC offset, and the alternative (.Local) would make a local build and the CI build disagree.
• Only the 15 most recent YouTube uploads are exposed by the feed; older videos drop off /videos on their own once newer ones push them past 15.
• Every video on the channel appears, including older workshop recordings, because the page mirrors the channel rather than a curated list.

If a feed is down at build time

• YouTube unreachable or empty → the build fails on purpose (see the errorf in layouts/_default/videos.html). Because the build fails, the deploy is skipped and the currently live gallery is left untouched, rather than being replaced by an empty page.
• Ghost unreachable → the build succeeds and the Blog card falls back to the last-known content baked into layouts/partials/feeds/ghost-latest.html. Update that fallback if the most recent post changes and you want the safety net to match.