You need a laptop with Node 22.13 or newer and pnpm, a room with Wi-Fi, and phones that can reach the laptop. Install it, start it, and open the admin page:

git clone https://github.com/adamXbot/OmegaQuiz.git
cd OmegaQuiz
npm install -g pnpm              # any recent pnpm; it switches itself to the pinned version
pnpm install --frozen-lockfile
pnpm start

The terminal prints two sign-in links, one for the admin page and one for the board. Open the admin link, press Sign in, and load a sample pack from Questions → Browse sample packs. Open the board link on the projector laptop, and have a phone on the same Wi-Fi scan the QR code it shows. That is a complete session, end to end, in a few minutes.

Choose what you want to do next:

Nothing leaves your server: there is no database, no third-party service and no telemetry. One Node process runs one game at a time and keeps its data in one folder on disk.

A session is 10 to 15 questions for a room of 5 to 100 people. Everyone starts in the race; a wrong answer (or no answer) knocks you out of the prize race, but you can keep answering for the engagement and accuracy stats. Whoever is still in at the end wins; a tiebreaker round settles it if more than one person is.

The flow on the day:

1. Open the lobby. The board shows a QR code, a six-digit join code and the address to type.
2. People join on their phones with their name and work email. Names appear on the board as they arrive.
3. Start the game. The first question goes to the board and every phone at once, with an optional countdown.
4. Each question: people tap an answer; the board shows how many have answered; you can offer a lifeline; when you are ready, Close & Reveal shows the right answer, how the room split, and the lesson behind it.
5. Next question, until the end, when the board runs a top-five ceremony and you download the results.

Four screens take part:

The app starts empty. Fill the bank in any of these ways:

• Browse sample packs. Three come bundled in samples/: a ten-question phishing-awareness round with five tiebreakers, plus shorter nature and pop-culture packs for a demo.
• Import a CSV or JSON file. Download the template from Admin → Questions, fill it in a spreadsheet, import it. JSON packs use the same format as the bundled ones, so a pack downloaded from GitHub loads as-is.
• Type them in. The editor has a card per question: the question, four options with the right one ticked, the lesson shown at the reveal, an optional image, a time limit, and presenter notes that only you see.
• Seed from a URL at deploy time, for a server that should come up with questions already in it.

Every card has a Preview button, and Preview all walks through the set, showing the real board and the real presenter view at projector size. It says whether a question fits at full size, had to shrink, or does not fit even at 60%, so a long lesson or a big image is never a surprise on the day. The editor warns under a card when something is long.

Packs keep several sets on the server. Save the current bank as a named pack, load another (it asks whether to switch the quiz title and tagline too), and Clear all questions when you want a clean slate. Before a load or a clear replaces the bank, the old bank is kept as an automatic snapshot in the same list, so neither can lose anything. Questions can only be changed while the lobby is open or after a game ends.

docs/QUESTIONS.md has the file format, the rules the editor applies, and the limits.

Timer. Each question stays open for a set number of seconds (45 by default, per question if you like). The board draws a ring, phones a bar, and the last five seconds go red. When time is up, answers lock, but nothing is revealed until you press Close & Reveal, so you can talk the room through it first.

Lifelines. 50/50 removes two wrong answers, Ask IT shows a hint, Skip reveals without scoring. Apply one yourself, or open a vote and let the phones decide. Each is one-shot by default; a setting refills them every question or at the tiebreaker.

Stopping or rewinding. From any question, reveal or the results screen: Back to lobby, keep players sends everyone to the lobby with the same join code, alive and on zero, so you can start over or swap the pack without sixty people re-scanning. Rewind to question N re-asks a question and undoes the scores and knock-outs from that point on. Both take two clicks, so a stray click never ends a live game.

Presenter view and clicker. Open /present on the facilitator laptop with the board on the projector as an extended display. It shows the current question with the right answer ticked and the lesson before the reveal, your notes for it, the live answer split, who has not answered, what comes next, and the controls. A presentation clicker (Logitech, Verbatim and most others) drives it: forward starts, closes and reveals, or moves on; back looks at past reveals on your screen only.

Sounds and theatre. Standard mode changes questions instantly with short cues on the board. Dramatic mode adds a suspense sting, a staged option reveal, a drum-roll before the answer and sounds on the phones. Each screen has a speaker button, which you can hide from the settings; M mutes the board either way.

docs/RUNNING-A-SESSION.md covers all of this in detail, including dropped phones and restarts.

Download results at the end of a game (or any time from Admin → Players) as Excel or CSV: one row per player with their score, whether they survived, and their answer and result for every question. The Excel file is colour-coded so a printout is scannable. CSV exports escape anything that a spreadsheet might try to run as a formula.

Every game is also kept in Admin → Sessions: when it ran, who played, who won, and a record of every answer and the questions asked. From there you can view the standings and a per-question breakdown, download the same Excel or CSV, load that game's questions to run them again, rename, or delete. Games stopped part-way are kept too, marked abandoned. Turn recording off in Branding if nothing should be kept.

docs/REPORTS.md lists every column and suggests a pass mark for training records.

There are no accounts and no passwords to manage. Two roles exist: host runs the game and opens the board and presenter view; admin does everything.

Magic links are the day-to-day way in. The boot banner prints one for each role, and Admin → Settings → Sign-in & keys mints more. Opening a link shows a page with a Sign in button; only that button signs the device in, so mail scanners and chat previews that fetch the link cannot use it up. A link is single-use and valid for an hour unless you choose otherwise: pick 10 minutes to 7 days, make it reusable until it expires to send a co-presenter, or let them scan the QR shown next to it. Revoke all links stops every unused one.

A sign-in lasts a week and extends every time you use it, and it survives a restart or a redeploy, so the facilitator laptop stays signed in between rehearsal and the event.

Recovery tokens (HOST_TOKEN, ADMIN_TOKEN) are the long-lived fallback for the form at /auth/login, for when every link has expired. Keep them in a password manager. Lost one? Rotate it from Settings, or with node server.js remint <role> on the server, without a restart and without signing anyone out.

Admin → Branding sets what people see: company name and logo, quiz title and tagline, the colour theme, the prize ladder (money or plain levels), and whether names are shown at the end. You can restrict joining to your company's email domain, write the privacy notice players agree to when they join, and add a call to action to the final screen ("Book time with IT").

The same page holds the game rules: the question timer, lifeline refills, whether answers lock in on the first tap or need a confirm, whether eliminated players can keep answering, the presentation style, the speaker buttons, and whether games are recorded. Everything here can also be set with an environment variable for a server that should come up configured; see .env.example.

For a room on one Wi-Fi network, the laptop running pnpm start is enough: phones use http://<your-laptop-ip>:3000. Set Public server URL in Branding to that address so the QR code and the board print it.

To host it on the internet it needs a long-lived process and a writable disk: Railway, Render, Fly.io, DigitalOcean App Platform and Docker all work. Vercel, Netlify and Cloudflare Pages will not, because serverless functions cannot hold a WebSocket connection or persistent storage. docs/DEPLOYMENT.md is the runbook: secrets, volumes, Cloudflare, checking it is healthy, and tearing down after an event.

Everything it keeps lives in one data folder (DATA_DIR, ./data by default): the question bank and saved packs, branding, the current join code, past games, sign-ins, and (optionally) the generated secrets. Back that folder up and you have backed up the app. Delete it and you have a fresh install.

The full table, with the error messages players see on their phones and what each one means, is in docs/RUNNING-A-SESSION.md.

Environment variables are documented in .env.example. The ones most people set:

CONTRIBUTING.md has the full guide, including what is in and out of scope. The whole app is one Node file and three plain HTML pages on purpose: no framework, no build step, no database. Run these before you open a pull request:

pnpm install --frozen-lockfile
pnpm audit --prod --audit-level=high
pnpm test

The Test workflow runs the same three commands on Node 22, 24 and 26 for every push and pull request, plus a Trivy scan of the Docker image, and repeats them nightly so advisory drift in pinned dependencies surfaces between PRs. Security reports go through SECURITY.md; release history is in CHANGELOG.md.

OmegaQuiz was written with the assistance of Claude (Anthropic), working alongside @AdamXweb. Every change was reviewed by a human before it landed, and the app was run in front of a real room before the features that came out of that were built.

Which commits are which is recorded in the history rather than asserted here:

git log --format='%an'                        # who authored each commit
git log --format='%b' | grep Co-Authored-By   # which were AI-assisted

MIT. Use it, fork it, ship it.

Not affiliated with, or endorsed by, the rights holders of Who Wants to Be a Millionaire? The gameplay format is inspired by the show (ITV / Sony Pictures Television); this implementation is original and shares no code or assets with it.