# Whole kits: original teaching tool

Fixed fictional recipe: standard uses 2 sheets / 3 minutes / 5 points; deluxe uses 3 sheets / 2 minutes / 7 points. Each count is an integer 0–20. Visitors change only sheet and minute allowances, integers 0–60. Exhaustive enumeration checks all 441 pairs using exact small integers. No dependency, HiGHS runtime, general model intake, autosave, upload or persistent browser storage.

JSON is UTF-8, at most 2048 bytes, with exactly four keys:

    {"format":"free4allhub-whole-kits","version":1,"sheets":10,"minutes":12}

JSON keys may be reordered; unknown keys, wrong versions, nonnumeric/fractional/nonfinite values and out-of-range allowances are rejected. Repeated JSON property names are rejected, including escaped duplicates. Imported strings are never interpreted as HTML or code. Review does not replace inputs; Cancel preserves them. Replacement clears results. Editing and export controls are locked while a file is reading or awaiting review. Cancel also cancels a pending read. Every edit clears results and pending replacement. CSV uses only fixed headers, validated integers and booleans; it includes all 441 combinations and both allowances. There are no visitor-authored text cells or formulas.

Zero kits always fits. “No kits fit” means no positive combination fits. All tied best rows would be returned and shown without an arbitrary tie-break. This fixed model has a unique optimum for every allowed allowance pair. No infeasible/unbounded status or solver status is implied.

Run `python3 scripts/build-hub.py` for the supported integration build, or `python3 scripts/build-whole-kits.py` for this route using the current family shell. Run `node --test tests/whole-kits.test.mjs` for logic/controller/source checks. Build output is allowlisted to this route; private research is not a build input.

Qualification: automated source and controller checks do not establish rendered-browser, native download, file-chooser, screen-reader, physical-phone or real-beginner acceptance. Those require separate verification of the current built bytes. Hosting is the supported module-loading route; offline file:// operation is not claimed. JSON is editable source for this exercise, not a general model or saved result. Never use this synthetic lesson to allocate money, staff, healthcare or opportunities.

Internal browser qualification (2026-11-12): the built route completed the 10/12 calculation, downloaded JSON, fresh-tab chooser review/cancel/reopen/explicit replacement, 8/12 revision, revised JSON download and another fresh-tab restore. The actual CSV was read back with 441 rows and sole best 4 standard / 0 deluxe. Zero allowance, decimal rejection, stale-result clearing and keyboard correction were checked. A 500-CSS-pixel browser window showed stacked controls with no horizontal overflow. This used the documented browser file-chooser API with actual download paths, not manual OS-picker operation. Physical phones, screen readers and real beginners were not tested. Malformed-file controls are automated, not browser acceptance.
