How jupyter/try-jupyter works
Project type and purpose
• This is a demo “Try Jupyter” full-stack front-end application with no traditional backend. It uses JupyterLite to embed a self-contained JupyterLab environment (Python, R kernels compiled to WebAssembly) entirely in the browser.
• Its purpose is to serve static assets (HTML, JS, CSS, notebooks, data) via GitHub Pages so that end users can run notebooks directly in-browser.
File structure
• Top-level config files
• .github/workflows/*.yml – CI workflows to build and deploy the static site
• .gitignore, build-environment.yml, environment-*.yml – environment configs
• readthedocs & conf.py – documentation build
• jupyter-lite.json, repl/jupyter-lite.json – JupyterLite front-end configuration
• content/
• data/ – sample datasets (CSV, GeoJSON, FASTA, Vega-Lite spec)
• notebooks/ – example notebooks (Intro, Lorenz, SQL, R)
• No server-side code: all logic lives in browser via WASM kernels
README for additional insights
• Confirms usage of JupyterLite and GitHub Pages
• Instructs users to edit notebooks via the live Try Jupyter page, then commit updates locally
• No mention of external APIs or microservices
a. Main components
• Static Web Host (GitHub Pages)
• Browser / Client
– JupyterLite Front-end (JupyterLab UI, JS bundle)
– Wasm Kernels (Python, R as WebAssembly modules)
– In-memory Virtual File System (Emscripten FS, persisted to IndexedDB)
– Sample content (notebooks & data) loaded at startup
• CI/CD Pipelines (GitHub Actions)
• Documentation Builder (ReadTheDocs)
b. Relationships & interactions
• Browser → (HTTP GET) → GitHub Pages: fetch HTML, JS, CSS, assets, content/
• JupyterLite UI loads kernels (WASM) and mounts the content folder into the in-memory FS
• User interacts through the notebook UI → code is executed by the Wasm kernel → reads/writes to FS → changes stored in IndexedDB
• GitHub Actions builds/validates the site on push and deploys to GitHub Pages
• ReadTheDocs uses conf.py to build project docs from content/ if documentation changes
c. Architectural patterns & principles
• Client-only architecture (no server backend)
• Static hosting + JAMstack approach
• WebAssembly for in-browser execution sandbox
• Virtual FS with client-side persistence
d. Key technologies
• JupyterLite (jupyter-lite.json)
• JupyterLab front-end (JS/CSS bundles)
• Emscripten FS + IndexedDB persistence
• GitHub Pages for static hosting
• GitHub Actions (deploy.yml, rtd-preview.yml)
Detailed but focused
• Every major moving part without splitting hairs on minor utilities