Examify

A single-screen exam practice demo. Shows one multiple-choice question at a time; after you answer, the correct choice is highlighted (and your wrong pick, if any). Click Următoarea → for the next one. Questions are shuffled per session and shown with no repeats; once you reach the end the deck reshuffles and loops — effectively infinite practice. Each question keeps its stable index (#id), shown in the header.

Run it

node server.js

Then open the printed URL:

• On this PC: http://localhost:8888
• From a phone/tablet on the same Wi-Fi: http://<this-PC-LAN-IP>:8888 (the server prints the exact address, e.g. http://192.168.x.x:8888)

The server is zero-dependency (Node built-ins only). On Windows, allow the firewall prompt for Node on private networks so other devices can connect. Change the port with PORT=3000 node server.js.

Open it via the server, not by double-clicking index.html — the app fetches questions.json over HTTP, which a file:// page can't do.

Deploy to GitHub Pages

The app is fully static and uses relative paths, so it runs on GitHub Pages as-is. A workflow at .github/workflows/deploy.yml publishes only the runtime files (index.html, styles.css, app.js, questions.json) — not mock.docx, server.js, or the extraction script.

One-time setup:

1. Create a repo on GitHub and push this folder:

   git remote add origin https://github.com/<you>/<repo>.git
   git push -u origin main
   

2. On GitHub: Settings → Pages → Build and deployment → Source: GitHub Actions.
3. Every push to main (or master) builds and deploys automatically. The live URL appears in the Actions run and under Settings → Pages, typically https://<you>.github.io/<repo>/.

server.js is only for local/LAN use; GitHub Pages doesn't run it.

Files

File	Purpose
index.html	The single screen
styles.css	Styling (card, choice states, responsive)
app.js	Quiz logic: load JSON, shuffle, reveal answer, loop
questions.json	245 questions: { id, exam, question, choices[], correctIndex }
server.js	Zero-dependency LAN static server

How the questions were extracted

The source is mock.docx. A .docx is a zip of XML with real text, so no OCR/vision is needed — _extract_docx.py reads word/document.xml directly. The correct answer in each question is the choice with a blue shading (w:shd fill 9EC4E7); other shading colors are stray noise.

The document holds three exam sections with inconsistent formatting (some choices are numbered lists, some are manual a)/b) text, some stems are glued to their first choice), so the parser groups by content: a stem ends in :/?, and the lines after it are its choices. Open-ended blocks and image captions are skipped, and Section C's duplicated questions are deduped. Each question gets a single stable global id (1–245) for display and for the non-repeat/loop logic; the source exam is kept in the exam field.

Regenerate questions.json from the docx with:

python _extract_docx.py