HealthAI Group Website

Public website for the Health AI Research (HealthAI) group at the Institute of Data Science, Maastricht University.

Architecture

Static site — no build step, no framework. Pages are plain HTML files that load structured content from content/*.yaml at runtime via cms.js. All content edits go in YAML; the HTML is not hand-edited for content.

File	Purpose
index.html	Homepage — hero, recent news, recent publications
people.html	Group members, filterable by theme
publications.html	Publication list with search, year slider, cite/BibTeX
research.html	Research themes and projects
news.html	News, blog posts, talks, podcasts — unified search/filter
students.html	Student projects (not linked from nav)
cms.js	YAML loader and shared rendering utilities
nav.js	Navigation bar and footer
styles.css	Global styles

Content files

File	What it drives
content/people.yaml	Group members — Faculty and Researchers sections
content/publications.generated.yaml	Publications (auto-generated — do not hand-edit)
content/research.yaml	Research themes and projects
content/news.yaml	News items, blog posts, talks, podcasts
content/site.yaml	Site-wide metadata
content/students.yaml	Student project listings

people.yaml schema

Each entry has: name, role, group, image_url, themes, bio, keywords, orcid, github, google_scholar, linkedin, personal_page, cris_url, work_url, email.

• group is either Faculty or Researchers
• On people.html, the Researchers group is split by role: entries with role: PhD Researcher appear under "PhD Researchers"; all others appear under "Postdocs, Researchers, and Research Software Engineers"
• github, google_scholar, linkedin store bare handles/IDs (not full URLs)
• cris_url, work_url, personal_page store full URLs
• image_url should be a local path (assets/people/<slug>.jpg) — never a hotlinked LinkedIn URL

news.yaml schema

The file has four top-level sections:

• news — day, month, year, title, body; optional url
• blog — managed by fetch_blog_posts.py; fields: title, url, date, author, excerpt, source
• talks — event, date, title, speaker, type; optional url (e.g. Zenodo DOI link), optional description
• podcasts — show, episode, title, guest, duration, date, icon

Scripts

All scripts are in scripts/ and require Python 3.

Add a new person

python3 scripts/add_people.py "Full Name" "https://www.maastrichtuniversity.nl/<profile-slug>"

Scrapes the UM staff page for bio, portrait photo, ORCID, and CRIS URL, then appends an entry to content/people.yaml. Review the entry afterward — the script defaults role to PhD Researcher; correct it if needed.

Publications pipeline

Run in order:

# 1. Fetch each person's ORCID works summary
python3 scripts/fetch_orcid_works.py [--refresh] [--name <substring>]

# 2. Fetch Crossref metadata for all unique DOIs
python3 scripts/fetch_crossref.py

# 3. Build publications.generated.yaml
python3 scripts/build_publications.py

Caches live in cache/orcid/, cache/crossref/, cache/abstracts/, and cache/summaries/. The pipeline is incremental — re-running only fetches new DOIs.

Optional: fetch full abstracts and generate plain-language summaries (requires Ollama running locally):

python3 scripts/fetch_abstracts.py
python3 scripts/generate_summaries.py

Sync blog feeds

python3 scripts/fetch_blog_posts.py [--dry-run]

Reads blog_feed URLs from content/people.yaml and syncs posts into the blog section of content/news.yaml. Manual entries (no source: feed field) are always preserved.

Enrich existing people entries

python3 scripts/enrich_people.py

Re-scrapes UM profile pages to refresh image_url and bio for existing entries. Caches HTML under cache/people_pages/.

Deployment

The site is a static file tree — deploy by pushing to the main branch. No build step required.

License

See LICENSE.