A browser-based game show simulator: thirteen contestants draft trivia
domains, duel head-to-head on a chess clock, and fight to become sole owner
of the board for a $100,000,000 grand prize. Each player is backed by a
local Ollama model that makes their live in-show decisions and answers
trivia; if Ollama isn't installed or a call fails, that player transparently
falls back to a scripted stand-in agent so the show never stalls (see
prototype/engine/ollama_agent.py).
A short demonstration video is available on Loom.
Dominion live
- Python 3.9+. No pip packages needed — the whole project is standard
library only (
requirements.txtexists but is intentionally empty, for tooling that expects one). - Ollama — optional. Powers the live agents; without it, every player just uses the scripted fallback.
python setup.py
Checks your Python version, and if Ollama is installed, pulls the models the
live agents use. Prints the exact command to start the server at the end —
use that command; it accounts for whether your system's Python 3 is called
python or python3. (If python setup.py itself doesn't run, try
python3 setup.py.)
python prototype/server.py
Then open http://localhost:8765 and click Start Show. Set
DOMINION_SCRIPTED_ONLY=1 first to skip live Ollama calls entirely and run
near-instantly, useful for quick local iteration.
See prototype/README.md for the engine layout,
live-agent details, and environment variables
(DOMINION_SCRIPTED_ONLY, DOMINION_PORT), and
design/Game Show Sim - Design Document.docx
for the full rule set and design rationale.
Everything above works as-is on a Mac — this section is just the same steps spelled out in full for anyone who's never used Terminal before.
- Get the project onto your Mac. However you received it (a zipped
folder, a
git clone), make sure it ends up somewhere easy to find, like your Desktop. - Install Python. macOS ships a stub
python3command tied to Xcode's developer tools, but it's often old and not meant for running real projects. Download the official installer from python.org/downloads (get 3.9 or newer, the latest release is fine), open the.pkg, and click through the installer with the default options. - Install Ollama (optional, but it's what makes the players "real").
Download the Mac version from
ollama.com/download, open the
.dmg, drag the Ollama app into Applications, then launch it once from Applications so it finishes installing its background service. Skipping this step is fine too — the show still runs end to end with every player on the scripted fallback agent instead of a live model. - Open Terminal. Press
Cmd+Space, typeTerminal, hit Enter. A Mac's Terminal is the same idea as Command Prompt/PowerShell on Windows, just a different app. - Navigate to the project folder. Type
cd(with a trailing space, don't hit Enter yet), then drag the project folder from Finder straight into the Terminal window — it'll paste in the full path automatically — and press Enter. - Run setup, then start the server:
(On a Mac,
python3 setup.py python3 server.pypython3is the right command, notpython, for most installs —setup.py's own output confirms which one to use if that guess is ever wrong.) - Open the show. Go to a browser and visit http://localhost:8765, then click Start Show.
Nothing here needs sudo or admin rights beyond the two .pkg/.dmg
installer clicks — this all runs as a completely ordinary local Python
process.
Built by Scott A. Cole, an AI strategy consultant and the author of 31 books on AI strategy, GenAI, and
decision-making, including the five-book Stop Learning AI series for executives who need to
make good AI decisions without becoming technical themselves. The app includes an About page and a Books page listing every title and edition (prototype/web/about.html, prototype/web/books.html). More projects, including Aegis Vector and Palimpsest, are at
github.com/ScottColeSW.
As an Amazon Associate I earn from qualifying purchases.
MIT — see LICENSE.