Fake Relics Obey Real Geology

Materials are assigned per component, and assignment is the first stage that consults the world beyond the object: what the region's geology offers, what the culture prefers, what its technology can work at this point in its history.

The catalogue

Sixteen materials, each a definition with tags, availability data and physical properties: bronze, iron, gold, silver, obsidian, flint, granite, jade, oak, ash, bone, antler, fired clay, glass, linen, leather. Tags group them (metal, stone, wood and so on) so that component compatibility can be expressed against kinds while selection lands on specifics.

Selection: two filters, a ladder, one product

1. Compatibility filter. Keep materials whose tags intersect the component's allowedMaterialTags; an empty constraint list admits everything.
2. Availability filter. Keep materials obtainable in this geology, locally or via a trade flow.
3. Fallback ladder. If availability empties the pool, fall back to the compatible set; if that's empty, the full catalogue. The function never hands weightedSelect an empty list, so it never throws on a populated catalogue.
4. Weighted draw. Survivors are weighted by the product culturalAffinity × phaseTechnology × scarcity.

Availability is a five-rung ladder per material per region: abundant, available, scarce, trade-only, absent. A trade-only material is rescued by a matching MaterialFlow in an active relationship window; an absent one is excluded outright. Scarcity then weights the survivors: 1.0, 0.6, 0.25, 0.15 down the rungs. A scarce material still appears, just rarely, and a trade material arrives at the doc's "present but uncommon" level.

// materials.ts: pinned by the calibration suite as ratios, not absolutes
const SCARCITY_WEIGHT: Record<string, number> = {
  'abundant':   1.0,
  'available':  0.6,
  'scarce':     0.25,
  'trade-only': 0.15,  // "present but uncommon" (doc 05 §7)
  'absent':     0,     // never reached: isAvailable excludes it first
};
// per surviving material:
// weight = culturalAffinity × phaseTechnology × SCARCITY_WEIGHT[level]

Provenance: the engine remembers, the player infers

Provenance answers a different question from findspot. Where an artefact was dug up belongs to stage 2 (the site, the layer, the deposition); where its raw material came from is decided here, at assignment time, because the answer depends on world state no later stage can reconstruct.

Trade windows close; a query run centuries later, against the world as it ended up, would get a different answer from the one that was true when the object was made. So the engine records the answer at the moment it's determined:

• Source: local, regional, trade or unknown. A mechanism, never an intent.
• Likely origin region: where the raw material plausibly came out of the ground.
• Trade-path reference: which relationship window carried it, when trade did.

All of it is hidden, and the hiding is the game design. A traded ingot, a looted trophy and a diplomatic gift can leave identical material signatures, so the player who sees a copper object from a culture with no copper deposits has to construct the explanation themselves; trade, migration, conquest, a stolen trophy and a misattributed culture are all live hypotheses, and the relationship data upstream genuinely supports several of them.

The engine's record exists so that a player's provenance claim can eventually be marked right or wrong against something.

Physical properties feed the next stage

Each material definition carries seven authored physical axes plus a keyed reactivity object, and each axis earns its slot by splitting a pair of materials that some downstream check must treat differently. An earlier model with a vague workable axis was retired for conflating brittleness, pliability and grain coarseness; these are the axes that replaced it, with the catalogue's own values.

One real entry, so the shape is concrete:

// data/materials.ts (comments abridged)
{
  id: 'obsidian', tags: ['stone'], craftDomain: 'stoneWorking',
  physicalProperties: {
    hardness: 5,      // Mohs 5–5.5: genuinely softer than flint, despite feeling sharper
    fragility: 6,     // textbook conchoidal fracture; flakes under a mis-strike
    rigidity: 7,
    grainFineness: 7, // a true volcanic glass: knapped edges sharper than steel
    porosity: 1,
    combustibility: 1,
    formability: 1,   // removal can't be steered into a raised form
  },
  reactivity: { oxidisation: -1 },
  decorability: { engravable: true, paintable: false, glazeable: false },
}

Obsidian's row is the axes doing their job in one material: its edge comes from fracture behaviour and grain, its hardness is unremarkable, and no single "workable" number could carry those three facts at once. Formability is the one axis read at the working state where the other six describe the finished object; fired clay is modelled wet there (6) and shatter-prone everywhere else (fragility 6), and the docs record that single-state inconsistency with the task id that owns resolving it.

Scaffolding, labelled

Milestone 2 runs this stage against mock world fixtures, and the module JSDoc records the consequences: there's no culture-to-region mapping yet, so availability answers "obtainable anywhere in this geology" rather than gating on the artefact's eventual findspot region, and the trade-path id is a synthesised string because relationships don't carry stable ids yet. Each shortcut names the roadmap task that retires it.