Agentic Learning AI Lab Website

Source for agenticlearning.ai.

Branches

• dev — development branch. Make changes here.
• main — production branch. Push/merge to main triggers CI deployment to GitHub Pages.

Prerequisites

• Node.js 20+
• Git LFS (PDFs and paper images are stored via LFS)
• Ghostscript for PDF compression (optional locally — CI handles it)

Image processing uses sharp (cross-platform, installed via npm).

Setup

git lfs install
git clone git@github.com:agentic-learning-ai-lab/agentic-learning-ai-lab.github.io.git
cd agentic-learning-ai-lab.github.io
npm install

Development

Start Tailwind in watch mode:

npm run start:tailwind

Tailwind classes are prefixed with tw- to differentiate from other classes.

Build

Full build pipeline:

npm run build

This runs the following steps in order:

1. build:tailwind — compile and minify TailwindCSS
2. generate_thumbnails — create 256x256 paper/people thumbnails via sips
3. generate_search_index — build searchable JSON index
4. build:arxiv — download arXiv HTML papers and images (skips existing)
5. build:compress — resize oversized images to max 1400px width
6. build:pages — render HTML pages from Handlebars templates

Build PDFs from LaTeX source

npm run build:arxiv:pdf        # compile PDFs + compress via Ghostscript

Force rebuild

npm run build:arxiv:force      # re-download all arXiv HTML and images
npm run build:compress:force   # re-compress all images

Adding a new paper

1. Add the paper entry to data/papers.yaml with arxiv, permalink, and optionally enable_full_paper: true
2. Add a paper image to assets/images/papers/
3. Run npm run build — this downloads the HTML, images, generates thumbnails, and builds pages
4. Optionally run npm run build:arxiv:pdf to compile a local PDF
5. Commit and push to dev, then merge to main to deploy

Large files (PDFs, paper images) are automatically tracked by Git LFS via .gitattributes.

Deployment

Deployment is automated via GitHub Actions. Pushing to main triggers the workflow:

1. Checks out the repo with LFS
2. Runs npm run build
3. Deploys to GitHub Pages via actions/deploy-pages

For manual deployment or local staging:

./deploy.sh staging    # build to staging/site/ for local testing
cd staging/site && python3 -m http.server 8000

Project structure

build/                     # Build scripts
  build_arxiv_papers.js    #   Download arXiv HTML, compile PDFs, compress
  build_pages.js           #   Handlebars template renderer
  compress_assets.js       #   Image compression (sips)
  generate_thumbnails.js   #   Thumbnail generation
  generate_search_index.js #   Search index builder
data/                      # YAML data files (papers, people, research areas)
research/                  # Per-paper directories (HTML, assets, PDFs)
assets/                    # Site-wide images, CSS, favicons
css/                       # Tailwind source and build output
includes/                  # Handlebars templates
.github/workflows/         # CI/CD (GitHub Actions)