Mars Open Facilities

The key facilities of the Mars colonies in the InterImm book, running live: propellant, power, air, iron, water and food. Each facility is a process model written for Homeostat, sized from published engineering numbers. A GitHub Actions workflow advances every plant to the present each hour and publishes its readings, and a static dashboard on GitHub Pages shows the status and the time series. Nothing runs on a server and nothing costs money.

The facilities

Facility	City	What it does	Design output
Procyon Propellant Works · 南河城推进剂工厂	Procyon, Isidis	Sabatier reactor and water electrolysis make CH4 and O2 for the shuttles	250 kg/h CH4, 1 t/h O2, 7.9 MW
Procyon Energy Station · 南河城能源站	Procyon, Isidis	Four 5 MWe heat-pipe fission modules and a 12 MWp solar field	about 24 MW on a sol average
Sirius Air Works · 天狼城空气工厂	Sirius, Isidis	Solid-oxide CO2 electrolysis (a scaled MOXIE) and the dome atmosphere loop	400 kg/h O2
Atlantis Ironworks · 亚特兰蒂斯城钢铁厂	Atlantis, Amazonis	Hydrogen direct reduction of hematite (as in HYBRIT) and an electric arc furnace	13 t/h steel
Horizon Starport Propellant Depot · 视界星港推进剂库	Horizon, Meridiani	LOX and liquid methane tank farms that load ships	4,100 t LOX, 1,140 t CH4
Trantor Ice Mine · 川陀城冰矿	Trantor, Hellas	Rodriguez wells melt buried glacial ice with hot water	21 m3/h water
Daoya Farm No. 3 · 稻芽三号农场	Terminus, Hellas	A 2 ha LED growth hall with CO2 enrichment	26 kg/h dry biomass while lit

Every facility page on the dashboard lists the sources behind its numbers; they are also in facilities/<id>/facility.yaml.

How it works

facilities/<id>/scenario.yaml   the Homeostat model: loops, instruments, plan, wear, maintenance, faults
facilities/<id>/facility.yaml   name, city, story, published tags, KPIs and sources
sim/update.py                   advances every plant to now and writes the data
sim/archive.py                  zips each finished month for a GitHub Release
web/                            the dashboard (plain HTML, CSS and JS; charts with uPlot)
.github/workflows/update.yml    runs sim/update.py hourly, publishes data, deploys Pages

• Story time. The site is set in the present of the InterImm book: 2219, the year terraforming begins. Every date is real time plus a fixed 70,491 days (STORY_SHIFT in sim/update.py), so 3 October 2026 reads as 3 October 2219; around leap years the story date can be a day behind. Epochs, file names, logs, releases and --now are all in story time. Each facility's since is its first year in service, from the book.
• Pseudo-live. Each plant's clock starts at its epoch (1 September 2219). A run advances it to the last whole 10 minutes before now. The engine is saved between runs (Engine.snapshot()), so the history is one continuous run.
• Deterministic. The same scenario and the same Homeostat and NumPy versions always give the same data, so if a saved engine cannot be used (a changed scenario, other versions), the plant is replayed from its epoch. requirements.txt pins the versions for that reason.
• Realistic, not real. Control loops, instrument lag, noise, quantization and drift, production plans, catalyst decay, fouling, valve wear, scheduled maintenance and random instrument faults all come from Homeostat. Sizes, yields and energy use come from the sources listed per facility.
• Operators' view. The data is what a plant historian would record: measured values, setpoints and controller outputs. Planned work is published 30 days ahead; faults are not announced, they only show in the data.

The data

The workflow keeps the current data on the data branch (one commit, replaced each run) and serves it with the site under data/:

File	Content
index.json	Every facility's status, mode and KPIs
<id>/meta.json	Description, tags, latest values, planned work, sources
<id>/recent.csv	The last 14 days, every 10 minutes
<id>/days/YYYY-MM-DD.csv	One file per UTC day, at least the last 60 days
<id>/sols.csv	One row per Mars Sol Date over the whole history (means, with n rows behind each)
<id>/log.json	Operations log: mode changes and maintenance
<id>/state/	The saved engine (not published on the site)
releases.json	Months already bundled into releases
timeline.json	The story-time shift and the epochs the data was made with; if either changes, the next run sets the old data aside and starts over

Every finished month of 10-minute data is also bundled into a zip and published as a GitHub Release tagged data-YYYY-MM (sim/archive.py). Daily files leave the data branch only after their month is released, so the full-resolution history is always in one place or the other.

Run it locally

python -m venv .venv && . .venv/bin/activate
pip install -r requirements.txt
python sim/update.py --data data            # advance to now (the first run compiles each plant, a few minutes)
python sim/update.py --data data --now 2219-09-10T00:00:00Z --only sirius-air
mkdir -p _site && cp -r web/. _site/ && cp -r data _site/data && python -m http.server -d _site

Adding a facility

Write facilities/<id>/scenario.yaml (see the Homeostat docs and the existing plants), check it with homeostat validate or a short run, then add facility.yaml with the tags to publish and the KPIs. The pull-request check runs every facility for two days.

Setup

GitHub Pages must use GitHub Actions as its source (Settings → Pages). The workflow needs no secrets.