Exclusivobeta
Odd One Out <mv-odd-one-out>
Detecção automática de valores atípicos em um conjunto de itens que deveriam ser iguais: desvio de configuração entre servidores, contêineres ou ambientes, inconsistências de preço e de estoque entre as variantes de um produto, um funcionário cujas permissões diferem das dos colegas com a mesma função, uma região cujas configurações destoam das demais. Envolva seus próprios cards ou linhas de tabela (os itens levam data-item e os valores data-field="key") ou passe um array items e ele renderiza uma tabela (um item por linha ou por coluna) ou cards. Para cada campo, ele encontra o valor majoritário (exato, sem diferenciar maiúsculas nem acentos, com sua própria função normalize ou comparator) ou, para números, a mediana com um teste MAD robusto ou uma tolerância absoluta / em %, e marca no lugar os valores que se destacam: um anel, uma célula colorida e um badge cujo glifo diz como ele difere (≠ differs, ▲ higher, ▼ lower, ∅ missing, + only here) com “differs from 7 of 8”. Campos em que cada item é diferente (nomes de host, IPs) ficam de fora, campos sem maioria clara são listados à parte, e itens aos quais falta um campo que a maioria dos outros tem também são sinalizados. Um resumo lista cada item atípico (“web-04 differs on 3 fields: TLS min version, Max body size, Region”) com seu valor atípico ao lado do majoritário, links diretos, navegação N / P entre as anomalias e uma ação “Align” opcional por valor, que passa por um evento mv-align cancelável para que o app faça a correção; um switch “Differences only” atenua tudo o que coincide (as visualizações renderizadas também recolhem os campos coincidentes). As verificações são refeitas ao vivo quando a marcação, os campos ou os dados mudam, e são anunciadas de forma discreta.
| Categoria | Exibição de dados |
|---|---|
| Tipo | Web Component (<mv-odd-one-out>) |
| Status | beta |
| Kit | Monitoramento sem fadiga de alertas |
| Também instala | button, switch, checkbox |
| Keywords | exclusive, culture, outlier, anomaly, drift, config-drift, consistency, majority, median, mad, compare, diff, audit, access-review, permissions, variants, inventory, fleet, devops, data-quality |
When to use
- A fleet of servers, containers or feature-flag environments must share one configuration and drift has to jump out
- A product's variants (sizes, colors, bundles) should share price, weight or tax codes except for the one mistyped
- An access review compares people in the same role and must surface the one with an extra or missing permission
- A small set of regions, stores or tenants should be configured alike and a reviewer wants the exceptions first
Avoid when
- Two or three documents of different kinds must be reconciled field by field against a reference, with a verdict → use Papers instead
- Several sources describe the same event over time and the disagreements are about what happened when → use Accounts instead
- Thousands of records are involved: compute outliers server-side and show them in a paginated table → use Data table instead
Instalação
node scripts/add.mjs odd-one-out --out ./src/marvelousAgente de IA com o servidor MCP do Marvelous UI: install_components({ slugs: ["odd-one-out"], target_dir: "<absolute path>/src/marvelous", framework: "react" }).
Arquivos copiados (dependências incluídas): tokens/tokens.css, core/base.css, core/dom.js, core/element.js, core/motion.js, components/odd-one-out/odd-one-out.js, components/odd-one-out/odd-one-out.css, components/button/button.css, components/switch/switch.css, components/checkbox/checkbox.css.
Uso
Início rápido, a menor marcação que funciona:
<mv-odd-one-out item-key="host">
<p data-item="web-01"><b data-field="host">web-01</b> <span data-field="tls">TLS 1.3</span></p>
<p data-item="web-02"><b data-field="host">web-02</b> <span data-field="tls">TLS 1.2</span></p>
<p data-item="web-03"><b data-field="host">web-03</b> <span data-field="tls">TLS 1.3</span></p>
</mv-odd-one-out>Marcação de referência, para usar como ponto de partida e personalizar com atributos, data-* e variáveis CSS:
<div id="ooo-demo" style="width:min(100%,74rem);margin-inline:auto">
<style>
#ooo-demo { display:grid; grid-template-columns:minmax(0,1fr); gap:1rem }
#ooo-demo .ooo-top { display:flex; align-items:flex-end; justify-content:space-between; gap:.75rem 1.5rem; flex-wrap:wrap }
#ooo-demo .ooo-top h3 { margin:0; font-size:1.125rem; letter-spacing:-.01em }
#ooo-demo .ooo-top p { margin:.25rem 0 0; color:var(--mv-fg-muted); font-size:.8125rem }
#ooo-demo .ooo-panel[hidden] { display:none }
#ooo-demo .ooo-fleet { overflow:auto; border:1px solid var(--mv-border); border-radius:var(--mv-radius-xl); background:var(--mv-surface); box-shadow:var(--mv-shadow-xs) }
#ooo-demo .ooo-fleet table { width:100%; border-collapse:separate; border-spacing:0; font-size:.8125rem; font-variant-numeric:tabular-nums }
#ooo-demo .ooo-fleet th, #ooo-demo .ooo-fleet td { padding:.5625rem .75rem; border-bottom:1px solid var(--mv-border); text-align:start; white-space:nowrap }
#ooo-demo .ooo-fleet tbody tr:last-child > * { border-bottom:0 }
#ooo-demo .ooo-fleet thead th { background:var(--mv-bg-subtle); color:var(--mv-fg-muted); font-size:.6875rem; font-weight:600; letter-spacing:.04em; text-transform:uppercase }
#ooo-demo .ooo-fleet td:first-child { font-family:var(--mv-font-mono); font-weight:600 }
#ooo-demo .ooo-fleet .ooo-dot { display:inline-block; width:.5rem; height:.5rem; margin-inline-end:.5rem; border-radius:50%; background:var(--mv-success); vertical-align:.05em }
#ooo-demo .ooo-controls { display:flex; align-items:center; gap:.75rem 1rem; flex-wrap:wrap; padding:.75rem 1rem; border:1px solid var(--mv-border); border-radius:var(--mv-radius-xl); background:var(--mv-bg-subtle) }
#ooo-demo .ooo-controls p { flex:1 1 18rem; margin:0; color:var(--mv-fg-muted); font-size:.8125rem }
#ooo-demo .ooo-log { flex:1 1 100%; min-height:1.1rem; margin:0; color:var(--mv-fg-subtle); font:.75rem/1.4 var(--mv-font-mono); overflow:hidden; text-overflow:ellipsis; white-space:nowrap }
</style>
<div class="ooo-top">
<div>
<h3 id="ooo-title">Production web tier · 8 hosts</h3>
<p id="ooo-sub">Every host should run the same edge configuration. Press N / P to walk through what differs.</p>
</div>
<mv-segmented id="ooo-scenario" value="fleet" label="Scenario">
<button value="fleet">Server fleet</button>
<button value="variants">Product variants</button>
<button value="access">Access review</button>
</mv-segmented>
</div>
<!-- Markup mode: the app's own table, rows marked with data-item, cells with data-field -->
<section class="ooo-panel" data-panel="fleet">
<mv-odd-one-out id="ooo-fleet" item-key="host" alignable label="Production web tier">
<div class="ooo-fleet">
<table>
<thead>
<tr>
<th data-field="host">Host</th><th data-field="image">Image</th><th data-field="tls">TLS min</th><th data-field="body">Max body</th>
<th data-field="workers">Workers</th><th data-field="keepalive">Keepalive</th><th data-field="gzip">Gzip</th><th data-field="region">Region</th><th data-field="p95">p95 latency</th>
</tr>
</thead>
<tbody>
<tr data-item="web-01"><td data-field="host"><span class="ooo-dot" aria-hidden="true"></span>web-01</td><td data-field="image">nginx 1.27.2</td><td data-field="tls">TLS 1.3</td><td data-field="body">10m</td><td data-field="workers">4096</td><td data-field="keepalive">65s</td><td data-field="gzip">on</td><td data-field="region">us-east-1</td><td data-field="p95">12 ms</td></tr>
<tr data-item="web-02"><td data-field="host"><span class="ooo-dot" aria-hidden="true"></span>web-02</td><td data-field="image">nginx 1.27.2</td><td data-field="tls">TLS 1.3</td><td data-field="body">10m</td><td data-field="workers">4096</td><td data-field="keepalive" id="ooo-keepalive">65s</td><td data-field="gzip">on</td><td data-field="region">us-east-1</td><td data-field="p95">14 ms</td></tr>
<tr data-item="web-03"><td data-field="host"><span class="ooo-dot" aria-hidden="true"></span>web-03</td><td data-field="image">nginx 1.27.2</td><td data-field="tls">TLS 1.3</td><td data-field="body">10m</td><td data-field="workers">4096</td><td data-field="keepalive">65s</td><td data-field="gzip">on</td><td data-field="region">us-east-1</td><td data-field="p95">11 ms</td></tr>
<tr data-item="web-04"><td data-field="host"><span class="ooo-dot" aria-hidden="true"></span>web-04</td><td data-field="image">nginx 1.27.2</td><td data-field="tls">TLS 1.2</td><td data-field="body">1m</td><td data-field="workers">4096</td><td data-field="keepalive">65s</td><td data-field="gzip">on</td><td data-field="region">us-west-2</td><td data-field="p95">13 ms</td></tr>
<tr data-item="web-05"><td data-field="host"><span class="ooo-dot" aria-hidden="true"></span>web-05</td><td data-field="image">nginx 1.27.2</td><td data-field="tls">TLS 1.3</td><td data-field="body">10m</td><td data-field="workers">4096</td><td data-field="keepalive">65s</td><td data-field="gzip">on</td><td data-field="region">us-east-1</td><td data-field="p95">12 ms</td></tr>
<tr data-item="web-06"><td data-field="host"><span class="ooo-dot" aria-hidden="true"></span>web-06</td><td data-field="image">nginx 1.27.2</td><td data-field="tls">TLS 1.3</td><td data-field="body">10m</td><td data-field="workers">4096</td><td data-field="keepalive">65s</td><td data-field="gzip">on</td><td data-field="region">us-east-1</td><td data-field="p95">15 ms</td></tr>
<tr data-item="web-07"><td data-field="host"><span class="ooo-dot" aria-hidden="true"></span>web-07</td><td data-field="image">nginx 1.27.1</td><td data-field="tls">TLS 1.3</td><td data-field="body">10m</td><td data-field="workers">4096</td><td data-field="keepalive">65s</td><td data-field="gzip">on</td><td data-field="region">us-east-1</td><td data-field="p95">48 ms</td></tr>
<tr data-item="web-08"><td data-field="host"><span class="ooo-dot" aria-hidden="true"></span>web-08</td><td data-field="image">nginx 1.27.2</td><td data-field="tls">TLS 1.3</td><td data-field="body">10m</td><td data-field="workers">4096</td><td data-field="keepalive">65s</td><td data-field="gzip">on</td><td data-field="region">us-east-1</td><td data-field="p95">13 ms</td></tr>
</tbody>
</table>
</div>
</mv-odd-one-out>
</section>
<!-- Data mode, cards: one product's variants -->
<section class="ooo-panel" data-panel="variants" hidden>
<mv-odd-one-out id="ooo-variants" item-key="variant" layout="cards" alignable label="Ridge Shell Jacket variants"></mv-odd-one-out>
</section>
<!-- Data mode, one column per person: peers in the same role -->
<section class="ooo-panel" data-panel="access" hidden>
<mv-odd-one-out id="ooo-access" item-key="name" layout="columns" alignable label="Support engineer, tier 2 permissions"></mv-odd-one-out>
</section>
<div class="ooo-controls">
<p id="ooo-hint">Align asks the app to apply the majority value; the check re-runs on its own when the table changes.</p>
<button type="button" class="mv-button" data-variant="outline" data-size="sm" id="ooo-drift">Simulate a config push on web-02</button>
<button type="button" class="mv-button" data-variant="ghost" data-size="sm" id="ooo-reset">Reset</button>
<p class="ooo-log" id="ooo-log" aria-live="off"></p>
</div>
<script type="module">
const root = document.getElementById("ooo-demo");
const fleet = document.getElementById("ooo-fleet");
const variants = document.getElementById("ooo-variants");
const access = document.getElementById("ooo-access");
const log = document.getElementById("ooo-log");
const say = (text) => { log.textContent = `${new Date().toLocaleTimeString("en-US", { hour: "numeric", minute: "2-digit", second: "2-digit" })} · ${text}`; };
const fleetMarkup = fleet.querySelector("tbody").cloneNode(true);
const variantData = () => [
{ variant: "Slate · XS", sku: "RSJ-SL-XS", price: "$189.00", weight: "310 g", material: "Recycled nylon", hs: "6201.40", rating: "20,000 mm", care: "Machine wash 30°C" },
{ variant: "Slate · S", sku: "RSJ-SL-S", price: "$189.00", weight: "325 g", material: "Recycled nylon", hs: "6201.40", rating: "20,000 mm", care: "Machine wash 30°C" },
{ variant: "Slate · M", sku: "RSJ-SL-M", price: "$189.00", weight: "340 g", material: "Recycled Nylon", hs: "6201.40", rating: "20,000 mm", care: "Machine wash 30°C" },
{ variant: "Slate · L", sku: "RSJ-SL-L", price: "$18.90", weight: "355 g", material: "Recycled nylon", hs: "6201.40", rating: "20,000 mm", care: "Machine wash 30°C" },
{ variant: "Slate · XL", sku: "RSJ-SL-XL", price: "$189.00", weight: "370 g", material: "Recycled nylon", hs: "6202.40", rating: "20,000 mm", care: "Machine wash 30°C" },
{ variant: "Slate · XXL", sku: "RSJ-SL-XXL", price: "$189.00", weight: "385 g", material: "Recycled nylon", hs: "6201.40", rating: "20,000 mm" },
];
const variantFields = [
{ key: "sku", label: "SKU" },
{ key: "price", label: "Price" },
{ key: "weight", label: "Weight" },
{ key: "material", label: "Shell", compare: "fold" },
{ key: "hs", label: "HS code", compare: "exact" },
{ key: "rating", label: "Waterproofing" },
{ key: "care", label: "Care label" },
];
const accessData = () => [
{ name: "Aisha Khan", tickets: "Read & write", pii: "Masked", refunds: "Up to $500", billing: "No", export: "No", logs: "Read", sso: "support-t2", mfa: "Hardware key" },
{ name: "Mateo Rossi", tickets: "Read & write", pii: "Masked", refunds: "Up to $500", billing: "No", export: "No", logs: "Read", sso: "support-t3", mfa: "Hardware key" },
{ name: "Yuki Tanaka", tickets: "Read & write", pii: "Masked", refunds: "Up to $500", billing: "No", export: "No", logs: "Read", sso: "support-t2", mfa: "SMS code" },
{ name: "Lars Nilsson", tickets: "Read & write", pii: "Masked", refunds: "Up to $500", billing: "Yes", export: "No", logs: "Read", sso: "support-t2", mfa: "Hardware key" },
{ name: "Priya Raman", tickets: "Read & write", pii: "Masked", refunds: "Up to $500", billing: "No", export: "No", logs: "Read", sso: "support-t2", mfa: "Hardware key" },
{ name: "Tomás Silva", tickets: "Read & write", pii: "Masked", refunds: "Up to $500", billing: "No", export: "No", logs: "Read", sso: "support-t2", mfa: "Hardware key" },
];
const accessFields = [
{ key: "name", label: "Name" },
{ key: "tickets", label: "Tickets" },
{ key: "pii", label: "Customer PII" },
{ key: "refunds", label: "Refunds" },
{ key: "billing", label: "Billing admin" },
{ key: "export", label: "Data export" },
{ key: "logs", label: "Production logs" },
{ key: "sso", label: "SSO group" },
{ key: "mfa", label: "MFA method" },
];
variants.fields = variantFields;
variants.items = variantData();
access.fields = accessFields;
access.items = accessData();
const copy = {
fleet: ["Production web tier · 8 hosts", "Every host should run the same edge configuration. Press N / P to walk through what differs."],
variants: ["Ridge Shell Jacket · Slate · 6 sizes", "Sizes share price, codes and labels; weight varies by size and is judged against the median, not exact equality."],
access: ["Support engineer, tier 2 · 6 people", "Same role, same access expected. One column per person, one row per permission."],
};
document.getElementById("ooo-scenario").addEventListener("mv-change", (e) => {
const v = e.detail.value;
for (const p of root.querySelectorAll(".ooo-panel")) p.hidden = p.dataset.panel !== v;
document.getElementById("ooo-title").textContent = copy[v][0];
document.getElementById("ooo-sub").textContent = copy[v][1];
document.getElementById("ooo-drift").hidden = v !== "fleet";
log.textContent = "";
});
for (const el of [fleet, variants, access]) {
el.addEventListener("mv-oddities", (e) => {
const { count, items } = e.detail;
say(count ? `mv-oddities · ${count} odd value${count === 1 ? "" : "s"} in ${items.join(", ")}` : "mv-oddities · everything agrees");
});
el.addEventListener("mv-align", (e) => {
const d = e.detail;
say(`mv-align · ${d.itemLabel} ${d.fieldLabel}: ${d.value} → ${d.majority ?? "(removed)"}`);
});
}
document.getElementById("ooo-drift").addEventListener("click", () => {
document.getElementById("ooo-keepalive").textContent = "75s";
});
document.getElementById("ooo-reset").addEventListener("click", () => {
fleet.querySelector("tbody").replaceWith(fleetMarkup.cloneNode(true));
variants.items = variantData();
access.items = accessData();
say("Reset.");
});
</script>
</div>Referência cultural
Vila Sésamo, Children's Television Workshop (Joan Ganz Cooney e Lloyd Morrisett) (1969, série). Em um quadro recorrente, quatro coisas aparecem lado a lado e o espectador precisa encontrar a que não é como as outras antes que a resposta seja revelada. Na interface, um conjunto de itens que deveriam ser idênticos é comparado campo a campo com a própria maioria, e os valores que quebram o padrão são apontados no lugar, contados e explicados, com uma forma de colocá-los de volta na linha.
API
Attributes
| Name | Tipo | Default | Description |
|---|---|---|---|
item-key | string (field key) | Field that names each item (host, sku, email). It is used as the item label and never compared. In data mode it is also the id property (default "id", then "name"). | |
compare | auto | exact | fold | number | number:<tol> | number:<tol>% | ignore | auto | Default rule for fields without their own (data-compare on a header or a field, or compare in the fields property). auto treats a field as numeric when every value parses as a number with the same prefix and unit ($, ms, GB), otherwise exact. exact compares trimmed text with collapsed spaces; fold ignores case and accents; number without tolerance flags robust outliers (modified z-score over the median absolute deviation, or any value off the common one when most items are identical); number:0.5 or number:5% flags values farther than that from the median; ignore never compares. Set exact on numeric-looking codes (HS codes, ZIPs, versions you do not want ranked) so they are reported as ≠ rather than ▲ / ▼. |
sensitivity | number | 3.5 | Modified z-score above which a numeric value is an outlier (Iglewicz and Hoaglin's 3.5). Lower flags more. |
quorum | number (0..1) | 0.5 | Share of items the majority must exceed. Below it the field has no clear majority: nothing is marked in place and the field is listed under “No clear majority”. A field where every item differs (hostnames, IPs) is treated as naturally unique and left alone. |
min-items | number | 3 | Fewest items needed to speak of a majority. With fewer, the summary says so and nothing is marked. |
differences-only | boolean | Dims every agreeing value (and agreeing items further) so the odd ones pop; rendered views also hide fields where every item agrees. Reflected by the summary's switch. | |
alignable | boolean | Adds an “Align” button to each odd value in the summary. It emits mv-align (cancelable); see the event for the default action. | |
summary | top | bottom | none | top | Where the summary panel goes, or none to only mark values in place (events, navigation methods and announcements still work). |
badges | auto | full | compact | none | auto | In-place badge text: full (“differs from 7 of 8”), compact (“7/8”, full text in the tooltip and for screen readers) or none (ring only). auto is compact inside table cells, full elsewhere. |
layout | rows | columns | cards | rows | Data mode only: a table with one item per row, a table with one item per column (best for many fields and few items, like environments or people), or a grid of cards. |
label | string | Consistency check | Accessible name of the rendered table and its scroll region. |
data-item | string (on an item) | Marks an item (card, row, section) and gives its id. Without any, direct children containing [data-field] are the items. data-item-label or a [data-item-title] child gives a nicer label. | |
data-field | string (on a value) | Key of a value inside an item. The value is data-value when present, the value of an input / select / textarea (checked state for checkboxes), otherwise the text (the component's own badges are skipped). An element with data-field outside any item is the field's header: its text is the field label and it receives an odd-count pill. | |
data-compare / data-field-label | string (on a header or value) | Per-field rule (same syntax as compare) and label. | |
data-odd-one-out / -kind / -item / -current / -count | set by the component | On values: odd | agree | split | neutral, and the kind (differs, higher, lower, missing, extra). On items: odd | agree. data-odd-one-out-current on the oddity being visited; data-odd-one-out-count on headers. The host gets data-state="odd | clear | few". |
Properties
| Name | Tipo | Description |
|---|---|---|
items | object[] | null | Data mode: one object per item; the component renders the view set by layout. null (default) reads the markup. Objects are copied; align changes the copies. |
fields | (string | { key, label?, compare?, format?, normalize? })[] | { [key]: { … } } | Field order, labels and rules. compare also accepts a function (a, b) => boolean; normalize(value) returns the text actually compared; format(value) the text displayed (data mode). Applies to markup mode too (labels and rules). |
oddities | object[] | Read-only: { id, item, itemLabel, element, field, fieldLabel, fieldElement, value, majority, agree, total, kind, basis ("majority" or "median"), badge, description } in item then field order. |
splits | { field, label, groups: { value, count }[] }[] | Read-only: fields without a clear majority. |
report | { field, label, status, majority, agree, total, rule }[] | Read-only verdict per field; status is agree, odd, split, unique, ignored or skipped. |
strings | object | Every visible and announced text (en-US by default), merged with the defaults. Plural entries are [one, other] arrays. |
Methods
| Name | Description |
|---|---|
next() / previous() | Moves to the next or previous oddity (wrapping): focuses it, scrolls it into view, marks it current and announces it. Returns false when there is none. |
focusOddity(index) | Visits a given oddity. |
align(item, field) | Same as the Align button: emits mv-align and applies the default action unless canceled. Returns true if not canceled. |
refresh() | Re-checks now (changes are otherwise picked up automatically, debounced). |
Events
| Name | Description |
|---|---|
mv-oddities | After the first check and whenever the oddities change. detail: { list (as the oddities property), items (odd item ids), fields (odd field keys), splits, count, total }. |
mv-align | Align was requested for an odd value. Cancelable. detail: { item, itemLabel, field, fieldLabel, value, majority (the target value; null for kind "extra", meaning remove), kind, element, fieldElement, data (the item object in data mode) }. Default action: data mode updates the item's copy and re-renders; markup mode sets the value of an input / select / textarea (firing input and change) or the element's text and data-value. Call preventDefault() to apply it yourself (framework state, API call); the component re-checks when your DOM or data changes. |
Content structure
| Name | Description |
|---|---|
(content) | Your items: a table with data-item rows, a grid of cards, sections… Never moved or rebuilt; the component only adds data-odd-one-out-* attributes, tabindex="-1" on a visited value, and small badges marked .mv-odd-one-out-own. |
CSS classes
| Name | Description |
|---|---|
mv-odd-one-out-summary | Summary panel (title, Differences only switch, previous / next, list of odd items, No clear majority line); data-state="odd | clear | few". |
mv-odd-one-out-badge | In-place badge appended inside an odd value (after it for inputs); data-kind; holds the glyph, the visible count and a screen-reader description. |
mv-odd-one-out-count | Odd-count pill appended to a field header. |
mv-odd-one-out-view / -table / -cards / -card | Views rendered in data mode. |
mv-odd-one-out-own | Every node the component creates; ignored when values are read. |
CSS variables
| Name | Default | Description |
|---|---|---|
--mv-odd-one-out-color | var(--mv-warning) | Tint of odd values, badges and item edges. |
--mv-odd-one-out-ink | color-mix(color, --mv-fg 48%) | Text color on odd tints (kept readable in light and dark). |
--mv-odd-one-out-agree-color | var(--mv-success) | Icon tint when everything agrees. |
--mv-odd-one-out-dim | 0.4 | Opacity of agreeing values under Differences only. |
--mv-odd-one-out-list-height | 22rem | Max height of the summary list before it scrolls. |
Accessibility
Color is never the only signal: every odd value gets a ring and a badge whose glyph encodes the kind (≠ differs, ▲ higher, ▼ lower, ∅ missing, + only here), and the badge carries a screen-reader description read with the value itself (“TLS 1.2. Differs from 7 of 8: TLS 1.2, majority TLS 1.3.”); for inputs the description is linked with aria-describedby instead. Field headers get an odd-count pill with a spoken equivalent. The summary is a labelled section whose heading states the result; odd items are a real list, each with a button jumping to the item and one per odd value whose accessible name includes the full description, plus labelled Align buttons (“Align TLS min version on web-04 to TLS 1.3”). N and P move to the next and previous oddity anywhere inside the component (ignored while typing in a text field or with modifiers; buttons with aria-keyshortcuts do the same): the value is focused (tabindex="-1" is added only while it is odd), scrolled into view, marked current with a thicker ring and announced (“Oddity 2 of 5: web-04, Region. Differs from 7 of 8…”). Changes after the first check are announced once in a polite live region with the new summary, and focus stays on the equivalent summary control after a re-render. The Differences only switch is a native checkbox with role="switch". The rendered table has real th headers with scope and a focusable, labelled scroll region. Reduced motion (prefers-reduced-motion or data-motion="reduce"): the one-shot highlight pulse and badge nudge are removed and scrolling is instant. Forced colors: odd values are outlined with the Mark system color and the current one with Highlight.