model.cjs is also embedded byte-for-byte in the standalone app. Node: require('./model.cjs'). Browser global: WildlifeCoDesign.

  • DEFAULTS = {target:110,maxSites:3,capitalLimit:null,landLimit:null}.
  • normalize(params) validates whole-number target 0–330, sites 0–3, optional capital ceiling in pounds (0–1,000,000,000) and land ceiling in square metres (0–1,000,000). null means no ceiling; zero means a zero ceiling. Reject unknown keys.
  • CATALOGUES: baseBridges, bridges, fences, monitoring; frozen plain objects.
  • compose('bridge-fence' | 'fence-monitor') returns {grouping,compatiblePairs,implementations} for the same fixed design problem, before query filtering. No intermediate Pareto pruning.
  • solve(params) returns normalized params, frontier, every query-feasible implementation in feasible, counts, rejection counts (overlapping), and the maximum capacity at the available number of sites.
  • frontier groups are {id,resources,implementations}. Equal resource vectors retain all implementation witnesses.
  • Each witness is {id,parts:{bridge,fence,monitoring},widths,sites,exec:{capacity},interfaces:{fenceRequired,fenceProvided,observationRequired,observationProvided},resources:{capital,land,annual},staffMilli}. Resources are pounds, square metres and pounds/year. staffMilli is thousandths of an FTE, already costed in annual expense.
  • replay(id,params) reconstructs from catalogue IDs, rechecks both interfaces and the request, and returns {valid,errors,implementation}. It does not trust exported resource values.

Catalogue and compatibility

Unordered bridge bundles of up to three components (including empty): widths 30/50/70m; fictional capacities 45/75/105 model crossings/day; capital £4,225,000/£6,255,000/£8,306,000; land 6000/10000/15000m²; upkeep £44,900/£65,500/£86,600 per year. A bundle of N bridges requires 2N guide-fence kilometres and 2N observation points.

Fence entries: none, or 2/4/6km × standard/durable mesh. Per kilometre: standard capital £100,000, land 500m², upkeep £2,000; durable capital £160,000, same land, upkeep £1,000. Fencing requires one observation point per two kilometres.

Monitoring entries: none, or 3/6/9 point capacities × field/assisted review. For p points: field capital 50000+30000p, staffing 100p milli-FTE, annual equipment 800p; assisted capital 120000+45000p, staffing 40p milli-FTE, annual equipment 2000+1400p. Both cost staff at £55 per milli-FTE per year (£55,000/FTE). Both use existing fence mounts and require no additional land in this model.

The only compatibility inequalities are B.fenceNeedKm <= F.km and B.observationNeed + F.observationNeed <= M.points. Empty B with excess F/M is allowed by these interfaces; the zero-work tuple dominates it at target zero. No hidden equal-width rule, no forced zero special case, no bridge history or time path.

Group (B,F) first, retaining the observation obligation, or group (F,M) first, exposing residual observation capacity M.points-F.observationNeed alongside F’s guide-km. Both constructions must return the identical complete witness set. Resources add componentwise.

The target selects provided capacity at least q. maxSites restricts implementation availability; it is not a fourth minimized resource. Capital/land ceilings restrict admissible resource vectors. All restrictions occur before minimization and are recomputed for every query.

Downloaded comparison record

The app’s wildlife-codesign-comparison.json contains model, request, selectedImplementation, replay, frontier, counts, units, boundary and the additive catalogues field. catalogues is the complete JSON representation of CATALOGUES: all three base bridges, 20 bridge bundles, seven fence options and seven monitoring options, including zero choices, interface quantities and component resources. Older records can lack this field; the model version remains wildlife-dpi-1 because its functions and numerical contract have not changed.

To check a saved witness without the live app, resolve its three parts IDs in the saved catalogues, reconstruct the bridge capacity and resources from its widths and baseBridges, check both compatibility inequalities and the saved request, then add bridge, fence and monitoring resources. Monitoring records include staffing and its annual cost breakdown. The saved catalogue makes these assumptions inspectable; it does not authenticate a record or turn synthetic coefficients into field evidence. At an infeasible request, the catalogues remain present while selectedImplementation and replay are null.

Run node export.test.cjs in this directory to exercise the production download handler with browser primitives stubbed, parse its actual JSON Blob and independently reconstruct saved witnesses and resource totals from the exported catalogues. This is a record-contract check, not a browser-download test.