# Browser acceptance — published tools Date: 2026-08-28 Release profile: `public_release` / `fc-3.4` Frozen source commit: `6e86ac1955a0c566c7aae521705e51371992ba8a` Canonical evidence identity: `7aa8588b50856816b7fce90dd6e743c26c6f16926071123291cb054352b6cd4a` This record covers the generator-integrated release bytes accepted before publication. The canonical viewer URL is [https://buddyog.github.io/topre-force-curves/](https://buddyog.github.io/topre-force-curves/), the public Open Graph image is `https://buddyog.github.io/topre-force-curves/assets/force-curve-bench-fc-3.4-og.png`, and the frozen publication tag is `fc-3.4`. The sections through **Responsive visual limitation** preserve the historical Force Curve Bench acceptance record. The historical EC Parts Builder `lib-6.0` acceptance and the current `lib-6.1` acceptance are recorded separately below. ## Interactive browser checks The generated single-file viewer was served from a local HTTP origin and tested in the Codex in-app Chromium browser at its available 1440 × 900 and 1280 × 720 desktop viewports. The final correction pass below used 1280 × 720. | Area | Result | Evidence checked | |---|---|---| | Fresh load | Pass | Empty chart, generated-fleet count, release identity, method, and footer rendered with no console warning or error. | | Single dome | Pass | Topre R2 45g showed both indices, the seven requested mechanical descriptors, detected force-wall onset as `3.96 mm` with no reference comparison, and recorded turnaround as the separate `4.25–4.26 mm` test limit. | | Shared displacement axis | Pass | Force view rendered one 0–4.5 mm fleet-wide scale and its visible force-wall marker. The 4.5 mm endpoint covers the released fleet maximum force-wall onset of 4.04 mm; recorded turnaround and longer samples do not expand the axis. | | Multi-selection | Pass | Topre R1 45g plus Sony BKE Brown 01/02/03 rendered together; physical siblings received distinct, stable semantic colors. | | Comparison modes | Pass | Force curves, Weight profile, Tactility profile, and Weight vs tactility each activated and exposed the correct chart label. | | Readout | Pass | Compact columns and More/Fewer columns worked; the expanded view included detected force-wall onset, turnaround, record ID, source set, and measurement cohort. | | Sidebar filtering | Pass | Matching nested children remained visible, nonmatching siblings stayed hidden, and selected nonmatching records remained available. | | Domes / Parts separation | Pass | Switching workspaces cleared incompatible selections; part assemblies displayed `Not calibrated` for both perception indices. | | Tests workspace | Pass | The 68 dome records and 8 part-assembly records remained separate, sortable/filterable populations. | | URL restoration | Pass | A four-record `?sel=` comparison restored the same four records in a fresh tab. | | Theme | Pass | Light and dark themes both rendered with the expected computed page colors. | | Fonts and requests | Pass | Inter and IBM Plex Mono loaded from embedded data; no third-party font request occurred. Runtime trace requests were limited to immutable raw CSV URLs at the frozen commit. | | PNG export | Pass | A fresh Topre R2 45g export was produced and visually inspected at 3005 × 1962 pixels: 611,063 bytes, SHA-256 `504f2e1a54a63ff58f1832627260231b816b0a82023d1e328bae45c6b882a224`. It included the shared 4.5 mm scale, the `3.96 mm` force-wall marker without a comparison, the separate recorded turnaround, the build/evidence footer, and a large centered `UNREAL KEYBOARDS` watermark at low opacity behind the chart. The watermark was absent from the live viewer. | | Rights/footer | Pass | Footer text is `© 2026 Brian “BuddyOG” Gebo — Unreal Keyboards. All rights reserved.` The separate build badge identifies the generated build. | ## Executable runtime batteries The final correction-focused Python/Node viewer and release suite passed all 157 checks. The batteries cover: - missing-network and malformed-success responses; - embedded offline fallback; - null and hostile-string behavior across cards, curves, profiles, scatter, readouts, tooltips, and PNG exports; - pointer-event and compact-layout behavior; - all four PNG export modes and evidence-stamped filenames; - the 390 px scatter/legend geometry and compact-chart height calculation; - review-versus-release profile separation; and - one shared fleet-wide axis that covers every released force-wall marker and is not expanded by recorded turnaround or longer samples; and - a PNG-only, low-opacity `UNREAL KEYBOARDS` watermark drawn behind the chart. Full generator parity also passed for all 184 semantic run bindings, with exact null/flag/audit agreement and a maximum scalar difference of `2.1316282072803006e-13` against the accepted JavaScript implementation. ## Responsive visual limitation The in-app browser advertised a viewport override but did not apply the requested 390 px width, even after reload and a fresh tab. Therefore this run does **not** claim a real-browser 390 px screenshot. The 390 px layout math, compact legend placement, touch/pointer paths, and responsive CSS contracts passed the executable batteries, and the supplied GUI itself was preserved as the accepted starting point. A final human phone-width spot check is still a useful deployment check, but this tooling limitation did not expose a viewer failure. ## Force Curve Bench fc-3.5 acceptance — 2026-09-04 This section records the viewer-only `fc-3.5` acceptance. The frozen raw-source commit, canonical evidence identity, 184 retained semantic run bindings, 180 unique acquisitions, full-precision metrics, 68-dome reference fleet, and `perception-rank-v1` values are unchanged from `fc-3.4`. Two independent release-profile generations produced the same 104 paths and were byte-identical. The generated release viewer SHA-256 is `108a6b3ee26b152909631a0dbada65bd3c2121f0421035294ef7055735c02f1f`. After this acceptance record was added, two complete packages were built from separate clean copies of predecessor commit `e9dde994e516ad01bfe384b4f5796508dff1c921` and compared path for path and byte for byte. Each package contains 825 regular files. The standalone package verifier passed both candidates. The complete generator suite passed **558 tests**, with **12 declared environment-dependent skips**. Its 39 focused fc-3.5 packaging tests include negative cases for stale or hand-edited staging, incorrect canonical URLs or tags, malformed component metadata, undeclared overlays, raw/canonical changes, pilot-checksum drift, line-ending drift, and protected Parts bytes. The canonical-evidence verifier independently passed all 108 declared checks. The release viewer was exercised in Chrome at 1440 × 900 and at a real 390 × 844 phone viewport. Browser acceptance covered: - a collapsed Filters section and separate collapsible Tests list on load; - flat, left-aligned search results and the Clear filters action; - Domes, Parts, and Tests workspaces; - Simplified default, Detailed mode, and label/visual controls; - Force curves, Weight profile, Tactility profile, Force-wall onset, and Weight-vs-tactility modes; - empty-selection automatic fleet loading plus Clear selection and Load all; - fixed-height comparison charts and single-selection profile views without a duplicate lower measurement strip; - force-wall group/reference presentation without recorded-turnaround substitution; - table sorting with horizontal scroll preserved (`2047 px` before and after); - direct selection restoration through `?sel=`; and - zero console errors, zero page errors, and no page-level horizontal overflow at 390 px (`scrollWidth = innerWidth = 390`). A real Simplified NiZ Purple 60g PNG export was generated and visually inspected at 2952 × 1966 pixels. It matched the on-screen measurements and 4.5 mm axis; used outside arrows with leader tails for the tight Drop dimension; included the detected force-wall marker, export date/time, build and dataset identity; and placed a faint centered `UNREAL KEYBOARDS` watermark behind the chart. The watermark remained absent from the interactive on-screen canvas. The fc-3.5 Open Graph image was also inspected at 2968 × 1928 pixels. Its SHA-256 is `2563d0fabbbd52e8610533d96c1e3d293f47db2fa9bc89f1131268ecd5e75bf0`. The viewer-only overlay preserved `dome-lab-parts.html` and `generated/packs/picker.staged.html` at the released `lib-6.1` SHA-256 `9fca852ce2a50a9c5aaa5dcd089c639846e41d8298dd90b4b597321e01c198ec`. The independent package audit also found every package CSV and every `canonical-evidence/` file byte-identical to the predecessor, with no deleted or undeclared path. `FORCE_CURVE_BENCH_RELEASE_MANIFEST_fc-3.5.json` records the new component boundary; the unversioned Force manifest remains the immutable `fc-3.4` record. ## Historical EC Parts Builder lib-6.0 acceptance The standalone release-profile builder passed 117 of 117 real-browser checks across 1280 px desktop, 390 px phone, and 320 px reflow viewports. It produced 12 inspected screenshots and recorded zero external requests, console errors, clipped controls, or horizontal-overflow failures. Its standalone runtime battery passed 45 of 45 checks, and the Shopify exact-origin bridge battery passed 12 of 12 checks. The complete generator suite passed 514 tests with 13 environment-dependent skips. Two release-profile generations were byte-identical, and the final site package passed its manifest, checksum, protected-predecessor, raw-data, and canonical-evidence verification. ## EC Parts Builder lib-6.1 acceptance The final `lib-6.1-review.1` prepublication build was reviewed in the browser and explicitly accepted by the owner. Its standalone runtime battery passed 85 of 85 checks. The complete generator suite passed 518 tests with 13 declared environment-dependent skips, and a clean second generation was byte-identical. The promoted `lib-6.1` release profile then passed 180 of 180 automated real-browser checks across 1280 px desktop, 390 px phone, and 320 px reflow viewports. It produced 12 screenshots, recorded no failed network requests, and passed the Shopify exact-origin bridge battery 12 of 12. Two independently generated release trees were byte-identical before packaging. Acceptance covered all 11 component rows; 28 keyboard starters; 4 manufacturer-parts shortcuts; all 10 recorded 2u assemblies; independent 1u and 2u ring-fit calculations; the DynaCaps supported-compression exception; measured-dome Force-Wall, Travel, and Dome compression displays; same- manufacturer fallback handling; and the part-scoped spring presentation. Deskeys and KLC springs produced one **Does not work** finding on the spring itself, MetaPulse produced one **Not verified** finding, unrelated rows were not blamed through duplicate pair messages, and genuine pair-specific findings remained active. The accepted review artifact's SHA-256 was `1f9150b6c1910ccef480dd3b561571e93ec7ca2da59f3fd273b6fa851193b546`. That hash records the prepublication artifact only. The promoted release bytes are identified by `SITE_RELEASE_MANIFEST.json` and `SHA256SUMS`. ## Deployment checks For the already published fc-3.4 release, load the canonical URL and confirm its Open Graph URL/image, open one shared `?sel=` URL, confirm the documentation and bundled-license links, and perform the human phone-width visual check described above. Also confirm that `dome-lab-parts.html` and the other protected continuity endpoints resolve, and spot-check one published raw CSV at its preserved relative path. For the published lib-6.1 builder, confirm that the standalone builder has no in-tool tabs or sibling-tool links; exposes 11 component rows, 28 keyboard starters, 4 manufacturer-parts shortcuts, 10 recorded 2u assemblies, and 68 exact specimen measurements; and uses **Works**, **Works with conditions**, **Does not work**, **Not verified**, **Manufacturer matched**, and dome-only **Not evaluated**. Confirm that measured-dome details and the visible Travel / Dome compression panel apply Force-Wall and ring-seat geometry as documented. The browser payload must contain 241 decision-changing pair edges after 97 generic matrix-expanded conical-spring edges are suppressed while the complete 3,570-edge audit closure remains intact. Selecting an affected spring must produce one finding and one badge on the spring row only; unrelated parts must remain unflagged, and the whole-build status must inherit the part issue. A genuine pair-specific condition or conflict must still render and affect the rollup. The Shopify wrapper owns Dome Lab navigation and must pass the iframe contract. `SITE_RELEASE_MANIFEST.json` and `SHA256SUMS` are the active site authority; the fc-3.4 manifest remains the protected predecessor record. ## Live deployment closeout — 2026-08-31 The publication checks above were completed against the public deployment: | Surface | Result | Public target | |---|---|---| | Dome Lab hub | Pass | | | Force Curve Bench page | Pass | | | EC Parts Builder page | Pass | | | Paper overview | Pass | | | Explanatory article | Pass | | | Standalone Force Curve Bench | Pass | | | Standalone EC Parts Builder (then `lib-6.0`) | Pass | | | fc-3.4 GitHub release | Pass | | | Historical lib-6.0 GitHub release | Pass | | | Preprint 2.0 DOI | Pass | | | Compendium DOI | Pass | | The storefront hub, paper page, article, and Tools navigation were inspected after publication. The standalone endpoints resolved to their released tools. Private review pages, the unpublished review theme, and the pre-release menu backup remain rollback assets rather than release authorities. ## EC Parts Builder lib-6.1 publication — 2026-09-04 The lib-6.1 update retained both public Parts Builder URLs and the Shopify embed contract. Git tag `ec-parts-lib-6.1` identifies the new builder release; `ec-parts-lib-6.0` remains available as the predecessor and rollback record. ## Beyond Snap Ratio Preprint 2.1 publication — 2026-09-04 The version-specific DOI resolved to the published 2.1 record with 21 files, a 4 September 2026 publication date, and the research compendium relation. The concept DOI resolved to the latest version; Preprint 2.0 remains available under its historical version DOI.