@escape-game-over/atlas 0.1.82 → 0.1.83

package/README.md

@@ -61,33 +61,37 @@ Linking to a page the active deployment switched off is a compile error, not a
 One file per deployment, saying only what differs:
 
 ```ts
-export const { project, redirects } = atlas.project(defaultMessages, defaultRoutes, {
-  url: "https://acme.example",
-  siteName: "Acme Rome",
-  icon,
-  themeColor: "#1d51e0",
-
-  enabledLocales: ["el-GR", "en-US"],
-  overrideRouting: { defaultLocale: "el-GR", prefixDefaultLocale: true },
-
-  overrideMessages: {
-    // Greek only — the English copy still falls through to the default catalog.
-    "home.hero.body": { "el-GR": "Η {company} κρατά το απόθεμα…" },
-  },
-  overrideRoutes: {
-    careers: { enabled: true },                                       // opt into a page
-    contact: { slugByLocale: { "el-GR": "epikoinoniste-mazi-mas" } }, // retranslate a URL
-  },
-})
+export const { project, redirects } = atlas.project(
+    defaultMessages,
+    defaultRoutes,
+    {
+        url: "https://acme.example",
+        siteName: "Acme Rome",
+        icon,
+        themeColor: "#1d51e0",
+
+        enabledLocales: ["el-GR", "en-US"],
+        overrideRouting: { defaultLocale: "el-GR", prefixDefaultLocale: true },
+
+        overrideMessages: {
+            // Greek only — the English copy still falls through to the default catalog.
+            "home.hero.body": { "el-GR": "Η {company} κρατά το απόθεμα…" },
+        },
+        overrideRoutes: {
+            careers: { enabled: true }, // opt into a page
+            contact: { slugByLocale: { "el-GR": "epikoinoniste-mazi-mas" } }, // retranslate a URL
+        },
+    }
+);
 ```
 
-Every field is named for what it *does* to the defaults, so a project file can
+Every field is named for what it _does_ to the defaults, so a project file can
 never be mistaken for the whole truth: `enabledLocales` picks from what the site
 declares, and every `override*` is a patch over it.
 
 ## The four steps
 
-Every step infers its types from the *values* of the last, so a consuming project
+Every step infers its types from the _values_ of the last, so a consuming project
 never writes a type argument:
 
 ```ts
@@ -111,29 +115,33 @@ Step 4 returns `t()`, `rich()`, `plain()`, `pathFor()`, `urlFor()`, `fileUrl()`,
 wired.
 
 Step 3's `redirects` is the one way to write a redirect: it types one
-deployment's old URLs against *that* deployment's pages and languages,
+deployment's old URLs against _that_ deployment's pages and languages,
 whichever project the site is currently built from. `siteRoutes()` takes what
 it returns and nothing else, resolves it, and writes `_redirects`:
 
 ```ts
 // projects/rome/redirects.ts
 export default redirects([
-  { from: "/jobs", to: { route: "careers" }, kind: "permanent" }, // ✗ if Rome has no careers page
-  { from: "/en/*", to: { route: "home", locale: "en-US", splat: true }, kind: "permanent" }, // /en/faq → /en-US/faq
-])
+    { from: "/jobs", to: { route: "careers" }, kind: "permanent" }, // ✗ if Rome has no careers page
+    {
+        from: "/en/*",
+        to: { route: "home", locale: "en-US", splat: true },
+        kind: "permanent",
+    }, // /en/faq → /en-US/faq
+]);
 ```
 
 `t()` is bound to a locale and knows each message's `{placeholders}` from its
 text:
 
 ```ts
-const t = site.translate(locale)
+const t = site.translate(locale);
 
-t("contact.intro")                                        // no placeholders, no second argument
-t("footer.copyright", { company: "Acme Rome", year: "2026" })
-t("footer.copyright")                                     // ✗ params are not optional
-t("footer.copyright", { company: "Acme Rome" })           // ✗ {year} missing
-t("footer.copyright", { company: "A", year: "1", x: "" }) // ✗ {x} is not a placeholder
+t("contact.intro"); // no placeholders, no second argument
+t("footer.copyright", { company: "Acme Rome", year: "2026" });
+t("footer.copyright"); // ✗ params are not optional
+t("footer.copyright", { company: "Acme Rome" }); // ✗ {year} missing
+t("footer.copyright", { company: "A", year: "1", x: "" }); // ✗ {x} is not a placeholder
 ```
 
 Nothing is auto-injected. Site-wide values are ordinary placeholders, passed
@@ -154,8 +162,8 @@ typed the English.
 ```
 
 ```ts
-const rich = site.rich(locale)
-rich("about.intro", { venue: "contact" })  // the destination, stated once
+const rich = site.rich(locale);
+rich("about.intro", { venue: "contact" }); // the destination, stated once
 ```
 
 | Mark              | Short for   | Is                                       |
@@ -235,7 +243,7 @@ regression-tested by an `@ts-expect-error` suite.
 | Two routes resolving to the same URL                                            | build throws, naming both                    |
 | A site icon that is not square, a multiple of 48px, and PNG                     | build throws                                 |
 
-The runtime throws are the ones types cannot see: facts about a *file* (an
+The runtime throws are the ones types cannot see: facts about a _file_ (an
 icon's real dimensions), facts only known after merging (URL collisions), or a
 value typed as plain `string` because it came from an env var. See
 [`docs/checks.md`](docs/checks.md) for the full split and how each error is
@@ -243,8 +251,8 @@ worded.
 
 ## Site-level versus per-page
 
-The same question answers it every time: *would this differ between two pages of
-the same site?*
+The same question answers it every time: _would this differ between two pages of
+the same site?_
 
 | Per page — passed to `metaFor()` | Site-level — declared in `atlas.project()` |
 | -------------------------------- | ------------------------------------------ |
@@ -262,7 +270,7 @@ actually renders, and why `twitter:card` is pinned.
 
 ## Config that tracks which pages exist
 
-A project switches routes off. Anything defined *alongside* a route then has two
+A project switches routes off. Anything defined _alongside_ a route then has two
 ways to rot: the page is built and its config is missing, or the page is gone and
 its config lingers. `PerRoute<typeof site, T>` rejects both.
 
@@ -274,7 +282,7 @@ against a known target gets it.
 const heroes = {
     home: { variant: "wide" },
     about: { variant: "tall" },
-} satisfies PerRoute<typeof site, Hero>;   // ✗ if `about` is off, ✗ if `contact` is on
+} satisfies PerRoute<typeof site, Hero>; // ✗ if `about` is off, ✗ if `contact` is on
 ```
 
 ## The rules this package keeps
@@ -290,9 +298,9 @@ const heroes = {
    runs with `"types": []` to keep it honest, and excludes `src/astro/` because
    that is the one place allowed to import the framework. Put something there
    only when it genuinely cannot work without Astro, and prefer moving the
-   framework-dependent *edge* there over moving the logic.
+   framework-dependent _edge_ there over moving the logic.
 4. **Infer from values, never from type arguments.** When a helper needs to be
-   typed against project data, take that data as a *parameter* and let TypeScript
+   typed against project data, take that data as a _parameter_ and let TypeScript
    infer it. This is what keeps rules 1 and 2 satisfiable without making callers
    write `<Locale, typeof baseMessages>`.
 
@@ -314,7 +322,7 @@ Whatever those entry points do not export is internal by construction rather tha
 by convention, which is what makes the file layout under `src/` a private detail:
 modules can be split, renamed or flattened without touching a consumer.
 
-The split *within* `astro/` is a further constraint, not tidiness. `./astro` is
+The split _within_ `astro/` is a further constraint, not tidiness. `./astro` is
 imported by an Astro config, which is evaluated before the build exists — so it
 may use `node:` built-ins but not `astro:assets`. `./astro/images` is the
 opposite: it runs inside the build, from a page, and is unusable from a config.
@@ -427,18 +435,18 @@ Three things the `.astro` setup works around rather than fixes. Each is open
 upstream, and each explains a rule that would otherwise read as an oversight.
 
 - [biomejs/biome#10321](https://github.com/biomejs/biome/issues/10321) —
-  *Astro: not formatting inside `{( … )}`* (`S-Bug-confirmed`). Biome formats
+  _Astro: not formatting inside `{( … )}`_ (`S-Bug-confirmed`). Biome formats
   `.astro` markup but not the interior of an expression, so the 31 `.map()`
   blocks across the templates keep their own indentation while the markup around
   them follows `indentWidth`. Hand-aligned until this lands.
 - [biomejs/biome#9944](https://github.com/biomejs/biome/issues/9944) —
-  *Astro: Parser error inside expression*. Biome parses the inside of `{ }` as
+  _Astro: Parser error inside expression_. Biome parses the inside of `{ }` as
   strict JSX, which rejects things Astro itself accepts. This is why every
   comment in a template is `{/* … */}` and never `<!-- … -->`: an HTML comment
   anywhere inside an expression is a parse error, and the JSX form is also
   stripped from the built HTML rather than shipped to readers.
 - [biomejs/biome#11275](https://github.com/biomejs/biome/issues/11275) —
-  *LSP formatting deletes Astro frontmatter/HTML in multi-project workspaces*
+  _LSP formatting deletes Astro frontmatter/HTML in multi-project workspaces_
   (`S-Bug-confirmed`). It names `html.experimentalFullSupportEnabled`, which this
   repository sets, and a workspace of several projects, which this repository is.
   The gate above is CLI-only and unaffected; editor format-on-save is the thing

package/docs/NOT-BUILT.md

@@ -4,7 +4,7 @@ Questions that come up, with the answer and the reasoning behind it. Every one
 of these looks like an obvious addition until you check — so the check is
 written down once, rather than redone by whoever thinks of it next.
 
-Each answer says what is *not* built and, where it matters, what is still fine
+Each answer says what is _not_ built and, where it matters, what is still fine
 to build. The two are easy to run together: "no star-rating markup" is not "no
 reviews", and reading it as the second would lose something worth having.
 
@@ -32,17 +32,17 @@ below still stands: the node renders nowhere, a model does not read it out of a
 `<script>` block, and it duplicates the page's own text at several KB per load.
 What is left in its favour is Bing, which has said its LLMs read schema.
 
-Which makes this a call a *site* makes per page, not one lib makes for every
+Which makes this a call a _site_ makes per page, not one lib makes for every
 repo it syncs to — hence a builder that has to be called and passed into a
 graph, rather than anything wired into `metaFor()`. A project that has not
 weighed the payload against a Copilot payoff it can actually measure should keep
 not calling it, which is the same answer this entry always gave.
 
-The two arguments that *sound* decisive and are wrong are recorded below, since
+The two arguments that _sound_ decisive and are wrong are recorded below, since
 they will be made again either way.
 
-Google restricted FAQ rich results to *"well-known, authoritative government and
-health websites"* in 2023, then [removed them from Search entirely](https://developers.google.com/search/docs/appearance/structured-data/faqpage)
+Google restricted FAQ rich results to _"well-known, authoritative government and
+health websites"_ in 2023, then [removed them from Search entirely](https://developers.google.com/search/docs/appearance/structured-data/faqpage)
 on 7 May 2026 and archived the documentation. A commercial site marking up its
 FAQ gets nothing rendered for the effort.
 
@@ -52,17 +52,17 @@ the HTML-to-text pipelines that feed a model routinely drop `<script>`, often th
 whole `<head>` with it. Where a block does survive, what arrives is
 `{"@type":"Question","name":…}` as tokens: the punctuation of a graph rather than
 a parsed one. Independent testing through 2025–26 has repeatedly found that
-content present *only* in structured data goes unextracted, while the same
+content present _only_ in structured data goes unextracted, while the same
 content in visible headings and paragraphs is read reliably.
 
 That is not an argument against the nodes lib does build. Schema is parsed at
-*indexing* time, so it reaches the AI surfaces built on top of a search index —
+_indexing_ time, so it reaches the AI surfaces built on top of a search index —
 AI Overviews, Copilot — even though a chatbot fetching the page directly sees
 only the visible text. It is an argument against reaching for this one expecting
 a model to pick the answers up, which is the reason it is given every time.
 
 The one cost that counts is the payload. Every other node lib builds is metadata
-*about* the page — an address, a price, a date. This one is a verbatim *copy of*
+_about_ the page — an address, a price, a date. This one is a verbatim _copy of_
 it: every question and every full answer, a second time, in a `<script>` block on
 the page already showing them. A ten-question FAQ is several KB duplicated on
 every load, and page weight is paid per visitor, forever.
@@ -116,12 +116,12 @@ answer for each is a fact about Google rather than a preference.
   clothes. Where there is a real number, `Offer` already carries it.
 - **`ImageObject`.** Both sibling implementations are `{"@type": "ImageObject",
   url}` and nothing else, which is a wrapper around the plain URL string Google
-  already accepts. Article documents the field as *"Repeated `ImageObject` or
-  `URL`"* — neither form preferred — and **width and height appear nowhere in
+  already accepts. Article documents the field as _"Repeated `ImageObject` or
+  `URL`"_ — neither form preferred — and **width and height appear nowhere in
   either the Article guidance or the image-licence guidance**. So having the
   dimensions is not a reason to emit the node; lib does have them for the share
   image, the icon and a video thumbnail, and Google has no use for them.
-  What the node *does* unlock is the **Licensable badge** in Google Images, and
+  What the node _does_ unlock is the **Licensable badge** in Google Images, and
   that needs `contentUrl` plus at least one of `license`, `creator`,
   `creditText` or `copyrightNotice` — rights data no project here holds. Build
   it the day someone wants that badge, and build it around the licence fields
@@ -137,8 +137,8 @@ does and does not buy.
 
 A section is the part of a publication a piece lives in: Sports, Business,
 Opinion. Open Graph's `article:section` takes one, and schema.org's
-`articleSection` is the same idea. It is *where a piece sits*, as against tags,
-which are *what it is about* — one home versus many labels.
+`articleSection` is the same idea. It is _where a piece sits_, as against tags,
+which are _what it is about_ — one home versus many labels.
 
 Everything here sits in the same place. A `section` field would say `"News"` on
 every post for as long as `/news` is the only feed, which is the URL repeated
@@ -172,7 +172,7 @@ Reviews a business collects about itself never qualify, so venue-level stars are
 unreachable however good the data is — they come from Google Business Profile
 instead, which is outside this repo.
 
-Room-level ratings on the `Product`/`Game` node *are* eligible. They need reviews
+Room-level ratings on the `Product`/`Game` node _are_ eligible. They need reviews
 collected on the site itself, shown in full on the page, and averaged over
 exactly the set displayed — never a number typed in beside a shorter list.
 
@@ -191,14 +191,14 @@ language registry](https://www.iana.org/assignments/language-subtag-registry) is
 not — which is the case `LanguageTag` in `config.ts` declines and says why.
 
 What moved is not the risk, it is the reason to hold the list at all. A country
-menu needs the codes *as data* — `countriesFor` renders and sorts them per
+menu needs the codes _as data_ — `countriesFor` renders and sorts them per
 locale — and once the list exists, deriving the type from it is free and the
 shape is strictly worse: it admitted 676 strings for 250 countries, including
 the withdrawn codes (`SU`, `YU`, `AN`) that look plausible enough to be typed by
 accident. So `CountryCode` is now `(typeof COUNTRY_CODES)[number]`, defined in
 `countries.ts` and re-exported from `contact.ts` beside `PostalAddress`.
 
-`CurrencyCode` stays a shape, because nothing needs a currency *list*: a
+`CurrencyCode` stays a shape, because nothing needs a currency _list_: a
 currency is written once per project, beside the prices it applies to, and a
 wrong one is visible the first time a page renders. Enumerate it the day
 something renders a currency picker, and not before.
@@ -221,7 +221,7 @@ without the rejection, but it means carrying a curated list for editor
 convenience alone, and a list that earns nothing else is a list that drifts
 unnoticed.
 
-The version worth building, if this ever bites: validate the *emitted* JSON-LD
+The version worth building, if this ever bites: validate the _emitted_ JSON-LD
 against a vocabulary dump, in a test. That sees all 800 types, and sees what no
 type could — a property that is not valid for the type it sits on.
 
@@ -245,12 +245,12 @@ subtitle: [
     { type: "bold", content: "hero.subtitle_text_2" },
     { type: "text", content: "hero.subtitle_text_3" },
     // …through _7
-]
+];
 ```
 
 **It fixes two things copy is not allowed to fix.** Word order, which differs
 per language — the array says the bold clause comes second, in every locale,
-forever. And the *number of runs*, which differs more: a clause English spends
+forever. And the _number of runs_, which differs more: a clause English spends
 four fragments on is one word in Greek, and there is no way to write that in an
 array whose length was decided by whoever typed the English. The extra keys get
 padded with empty strings or the spare ones go untranslated, and either way the
@@ -269,7 +269,7 @@ compared across locales by the machinery that already compares `{placeholders}`
 — see `Tokens` in `i18n/placeholders.ts` — so a translation that drops the link
 fails the build rather than the page. That check is the reason marks are safe to
 put in copy at all, and it is what the array shape genuinely did buy: with the
-structure in config, no locale *could* lose a link. Marks had to earn that back.
+structure in config, no locale _could_ lose a link. Marks had to earn that back.
 
 **A mark names a slot, never a destination**, and that distinction is the second
 thing the array shape got right and the first version of this got wrong. `venue`
@@ -294,7 +294,7 @@ whose whole claim is that a translation cannot silently go wrong. The migration
 is happening once; this is the moment it costs least to fix, and every store
 collapsed to marks is a store whose copy a translator can finally read.
 
-### Should lib carry the consent *banner* as well as the tags?
+### Should lib carry the consent _banner_ as well as the tags?
 
 **No — and the line between the two halves is the point.**
 
@@ -315,12 +315,12 @@ somebody's compliance.
 **One part of it is lib's, and is built as of August 2026: whether there should
 be a banner at all.** `consentRequired(analytics)`, with the reasoning in
 `src/analytics/index.ts`. It sits here rather than in each project because it is
-a question about *tags* — which lib owns — wearing the costume of a question
+a question about _tags_ — which lib owns — wearing the costume of a question
 about banners, and because the obvious check is wrong in both directions.
 Everything above about the banner itself still stands.
 
 The seam between the halves is a global — `CONSENT_UPDATE_GLOBAL` — because
-that is the only place they *can* meet: the head script runs before any module
+that is the only place they _can_ meet: the head script runs before any module
 exists. lib emits an updater covering exactly the signals it defaulted, so a
 banner never names them and the two lists cannot drift. `src/astro/consent.ts`
 carries the browser-side parts that are the same everywhere — reading an answer,
@@ -351,7 +351,7 @@ Two traps live with this, both spelled out where the code is. A preconnect is
 reused only by a request whose CORS mode matches it, so one marked
 `crossorigin` for a classic `<script src>` warms nothing while looking entirely
 correct — fonts are the opposite case, and the two must not be copied between.
-And Tag Manager's loader URL is written *by* the inline block, so nothing in the
+And Tag Manager's loader URL is written _by_ the inline block, so nothing in the
 markup names that origin until the script has run, which is what makes the
 connection worth stating rather than leaving to the preload scanner.
 
@@ -384,7 +384,7 @@ repo, which is why this entry exists — reading that file is what prompts the
 question.
 
 Nothing is served by Apache any more. Worth knowing the seam is already there if
-that ever changes: the site resolves rules to plain *data*, and
+that ever changes: the site resolves rules to plain _data_, and
 `buildCloudflareRedirects` is one renderer over them rather than the only
 possible one, so another host is a new function and no redesign. The two
 non-redirect halves of that `.htaccess` — extension resolution and slash
@@ -419,13 +419,13 @@ staleness, and `generate()` rebuilding per request does not cure it — it deriv
 the same answer from the same frozen inputs.
 
 Routes look like the fix, and they genuinely are one. Measured on the b2b
-example: an injected endpoint that reaches the site through the *page* graph
+example: an injected endpoint that reaches the site through the _page_ graph
 served the edited string immediately, side by side with the integration's
 middleware still serving the old one.
 
 **And `_redirects` can be one of those routes** — worth recording, because the
 old comment in `site-routes.ts` said it could not be, and that was wrong. Astro's
-page walker skips a *filename* beginning with `_`; an injected route names its
+page walker skips a _filename_ beginning with `_`; an injected route names its
 pattern and its entrypoint separately, so the underscore lands only on the URL.
 It was built, served in dev, and written to `dist/_redirects` by the route alone.
 
@@ -474,7 +474,7 @@ replaced.
 **No, and a post's trail should point at the list, not at the page it is listed
 on.**
 
-A `BreadcrumbList` states *hierarchical* position, and pagination is lateral
+A `BreadcrumbList` states _hierarchical_ position, and pagination is lateral
 movement between siblings at one level: `Home › News` is the trail on `/news`
 and on `/news/page/7` alike. Google renders breadcrumbs in results, so a
 trailing `News (2)` buys noise in the SERP for a listing page that is already a
@@ -505,7 +505,7 @@ looking at" better than any link can.
 With `prefixDefaultLocale: false` the default locale is served at `/about-us`
 and nothing is built under `/en-US/`. Claiming the pages under it one rule at a
 time does not work: slugs are translated per locale, so the URL a reader lands
-on by editing `/el-GR/sxetika-me-emas` down to English keeps the *Greek* slug —
+on by editing `/el-GR/sxetika-me-emas` down to English keeps the _Greek_ slug —
 `/en-US/sxetika-me-emas`. Covering that means the cross product of the prefix
 with every locale's spelling of every slug, a `_redirects` of hundreds against
 the 2000 a host reads.
@@ -521,7 +521,7 @@ left out if anything is built under the prefix.
 
 ## Adding to this
 
-One question, one answer, one reason. Say what is *not* built and what still is
+One question, one answer, one reason. Say what is _not_ built and what still is
 — the scope is the part that gets lost. If the answer rests on someone else's
 decision, put it in the first group and name the date, so a reader can tell a
 settled fact from one worth re-checking.

package/docs/checks.md

@@ -18,12 +18,12 @@ type-checking would pass all of it.
 
 ## Compile errors, and where they are stated
 
-Most guarantees are compile errors, stated in a *constraint* rather than in the
+Most guarantees are compile errors, stated in a _constraint_ rather than in the
 parameter type — in parameter position a single unrelated mistake widens the
 inferred type and every key then reports as malformed, burying the real error in
 a cascade.
 
-That rule inverts when the validator maps the *whole* object rather than one of
+That rule inverts when the validator maps the _whole_ object rather than one of
 its properties, as the slug-depth checks do: in the constraint, any unrelated
 malformation makes the argument itself unassignable and TypeScript reports at the
 call site instead of the offending line. Those two live in parameter position for
@@ -50,7 +50,7 @@ sees.
 | a `lastmod` that is not a `YYYY-MM-DD` calendar date                                      |                                                                         |
 | config for a route this project does not build, or missing for one it does (`PerRoute`)   |                                                                         |
 
-The runtime checks are the ones types cannot see: facts about a *file* (an icon's
+The runtime checks are the ones types cannot see: facts about a _file_ (an icon's
 real dimensions), facts only known after merging (URL collisions), or a value
 typed as plain `string` because it came from an env var, where the literal checks
 pass vacuously.
@@ -87,7 +87,7 @@ consuming project's own `type-tests/`, which is also the only place the
 augmentation is real. Both examples carry one:
 [`b2c`](../examples/b2c/type-tests/wiring.ts) asserts against a `public/` with
 two files in it, and [`b2b`](../examples/b2b/type-tests/wiring.ts) against one
-with none, where the union is empty and *every* path is rejected.
+with none, where the union is empty and _every_ path is rejected.
 
 Declared analytics events are the other augmentation, and the opposite case:
 nothing is inferred, so a declaration means the same thing here as in a project.
@@ -152,10 +152,10 @@ does not carry; see [`client-scripts.md`](client-scripts.md).
 
 ## What a consumer checks
 
-A consuming project keeps its own thin suite for *wiring* — that its catalog,
+A consuming project keeps its own thin suite for _wiring_ — that its catalog,
 routes and config actually satisfy these contracts, and that its own types have
 not been loosened by a stray `any` or a dropped `satisfies`. It does not restate
 the machinery's behaviour; that is what the two suites above are for.
 
 Both examples' `type-tests/wiring.ts` are deliberately short, and every assertion
-in them names something that must stay *rejected*.
+in them names something that must stay _rejected_.

package/docs/client-scripts.md

@@ -22,11 +22,11 @@ than a view of something else, and a `hidden` on a clear button, a no-results
 message or a section with nothing left in it is binary and derived — there is no
 version of "the clear button while the field is empty" that is a matter of taste.
 It sets no class, no style and no `aria`, which is where the choices live. It
-also never *finds* an element: it is handed the ones a project looked up, so no
+also never _finds_ an element: it is handed the ones a project looked up, so no
 attribute name in this package has to be matched by any consumer.
 
 `youtube` is the other exception, and the one that cannot be avoided: the player
-*is* the behaviour, so something has to put an iframe in the page. What it sets
+_is_ the behaviour, so something has to put an iframe in the page. What it sets
 on it has no taste in it — the address, the title, the permissions, full screen,
 the referrer policy — and everything that does, classes above all, comes from the
 caller through a callback. It too is handed the button and the box rather than
@@ -78,10 +78,10 @@ the options, and when one drifts nothing reports it — a listener added without
 its twin simply survives teardown.
 
 ```ts
-const detach = slider.attach(viewport);       // carousel: one element
-const detach = list.attach();                 // filters: no element, only popstate
-const detach = loop.attach(video);            // background-video: the <video>
-const detach = trailer.attach(button, box);   // youtube: what is pressed, what it replaces
+const detach = slider.attach(viewport); // carousel: one element
+const detach = list.attach(); // filters: no element, only popstate
+const detach = loop.attach(video); // background-video: the <video>
+const detach = trailer.attach(button, box); // youtube: what is pressed, what it replaces
 ```
 
 `element` says the same thing in the shape a custom element needs: its function
@@ -90,8 +90,10 @@ element leaves, and what it returns is the undo for everything else.
 
 ```ts
 export const thing = element("go-thing", ({ root, signal }) => {
-    ref(root, "[data-thing-button]").addEventListener("click", open, { signal });   // ends with the visit
-    return loop.attach(root);                                                        // and so does this
+    ref(root, "[data-thing-button]").addEventListener("click", open, {
+        signal,
+    }); // ends with the visit
+    return loop.attach(root); // and so does this
 });
 ```
 
@@ -129,7 +131,7 @@ const detach = list.attach();
 
 A field is its kind, and its URL parameter is its own name — to rename the
 parameter, rename the field. `{ kind: "text", url: false }` keeps a field out of
-the URL, for a list whose page owns the URL itself. `values` is typed *from*
+the URL, for a list whose page owns the URL itself. `values` is typed _from_
 `fields`, so a `flag` demands a boolean and a mismatch is a compile error rather
 than an item that silently never matches. It is read once per item, at
 construction. `matched` is a set of the caller's own items, so there is no key
@@ -149,7 +151,7 @@ and then answers to any of them:
 values: (room) => ({ category: data(room, "categories").split(" ") }),
 ```
 
-An escape room is adventure *and* sci-fi; a film is a comedy *and* a drama. This
+An escape room is adventure _and_ sci-fi; a film is a comedy _and_ a drama. This
 is a widening of `choice` rather than a fourth kind, because a kind for it would
 carry the same state, the same parameter, the same `set` semantics and the same
 control, and differ by one operator — two kinds doing one job, which is what the
@@ -158,7 +160,7 @@ bare string and is untouched.
 
 **`text` folds both sides** — `NFD`, drop the combining marks, lowercase, trim.
 This is the piece that is missing nearly everywhere, and it is not cosmetic: a
-reader who types `malaga` is otherwise told there is no *Málaga*, and every
+reader who types `malaga` is otherwise told there is no _Málaga_, and every
 Romanian entry spelled with `ă`, `ș` or `ț` hides from anyone whose keyboard does
 not carry them. Item text is folded once at construction; only the query is
 folded per render.
@@ -170,8 +172,8 @@ whole query as one substring makes that accident load-bearing, so a two-word
 query fails whenever the words are not adjacent.
 
 **`flag` is a narrowing, not a two-way filter.** Off keeps everything rather than
-keeping the items that are *not* flagged — "In stock", unchecked, does not mean
-"out of stock". On, off and *either* is a `choice` with two values; folding that
+keeping the items that are _not_ flagged — "In stock", unchecked, does not mean
+"out of stock". On, off and _either_ is a `choice` with two values; folding that
 into a toggle produces a control with a third position nothing can reach.
 
 Flags are not exclusive. A second toggle is another **field**, not another value
@@ -183,7 +185,7 @@ parameters or comma-joined) that nothing has needed yet.
 ### Facet counts: `matchedWithout`
 
 A dropdown that says how many results each option would give needs a different
-set from `matched`: every field *except its own*. Counted against `matched`, a
+set from `matched`: every field _except its own_. Counted against `matched`, a
 facet whose value is already picked can only ever offer that value — every other
 option reads zero, because the items holding them were just filtered out by the
 very field being asked about.
@@ -220,7 +222,7 @@ construction.
 1. **Replace while typing, push on a choice.** An input emits an event per
    keystroke, and history should not fill with a word being spelled; picking a
    category is one deliberate act the back button should undo. The default comes
-   from the kind, but it is really a fact about the *input* — a `<select>` or a
+   from the kind, but it is really a fact about the _input_ — a `<select>` or a
    preset link bound to a `text` field wants `list.set("q", v, { history: "push" })`.
 2. **Empty parameters are dropped**, and an empty state is the bare path rather
    than a trailing `?`. Both spellings are the same page; one of them is worth
@@ -232,7 +234,7 @@ construction.
    because nobody deleted it. Fields write in declaration order, so one state
    always produces one URL.
 
-`attach` reads the URL *before* the first `onChange`. That is what lets a list
+`attach` reads the URL _before_ the first `onChange`. That is what lets a list
 work before this script arrives: a `<form method="get">` submits, the server
 renders the filtered page, and the first announcement continues from that state
 instead of flashing the whole list back. Nothing writes to the URL before
@@ -263,7 +265,8 @@ byte.
 
 ```ts
 const list = filters({
-    fields, items,
+    fields,
+    items,
     onChange({ matched }) {
         // …everything this list draws for itself
     },
@@ -317,13 +320,13 @@ onChange({ matched }) {
 just written it and that is the shortest true description of what survived — and
 because reading it is what lets the helper nest. A country is empty when every
 city in it is hidden; a region is empty when every country in it is hidden,
-*including the ones the previous call just hid*.
+_including the ones the previous call just hid_.
 
 **Which makes the order load-bearing: innermost first.**
 
 ```ts
 hideEmpty(countries, (c) => citiesIn.get(c) ?? []);
-hideEmpty(regions,   (r) => countriesIn.get(r) ?? []);
+hideEmpty(regions, (r) => countriesIn.get(r) ?? []);
 ```
 
 Run the other way round, the regions are judged against countries nothing has
@@ -422,9 +425,20 @@ import AtlasElement from "@escape-game-over/atlas/astro/element";
 
 <script src="./faq.ts"></script>
 
-<AtlasElement of={faq} class="…">
-    <input data-faq-search type="search">
-    <details data-faq-row data-key={key}>…</details>
+<AtlasElement
+    of={faq}
+    class="…"
+>
+    <input
+        data-faq-search
+        type="search"
+    >
+    <details
+        data-faq-row
+        data-key={key}
+    >
+        …
+    </details>
 </AtlasElement>
 ```
 
@@ -456,11 +470,15 @@ attribute, and the script writes only that run:
 ```
 
 ```astro
-<RichText spans={rich("faq.matched", { count: "" })} attrs={{ count: { "data-faq-count": "" } }} />
+<RichText
+    spans={rich("faq.matched", { count: "" })}
+    attrs={{ count: { "data-faq-count": "" } }}
+/>
 ```
 
 ```ts
-for (const each of refs(root, "[data-faq-count]")) each.textContent = String(hits.size);
+for (const each of refs(root, "[data-faq-count]"))
+    each.textContent = String(hits.size);
 ```
 
 Marks are parsed before placeholders are filled, so a run filled with `""`
@@ -557,7 +575,7 @@ What it owns is what every hero loop needs and most get quietly wrong:
 - **The right cut after a rotation.** A browser chooses among `<source media>`
   once, when the element loads, and never again, so a phone turned sideways
   keeps the portrait cut. This listens to every source's query and calls
-  `load()` when the answer changes — *the answer*, not any crossing: a later
+  `load()` when the answer changes — _the answer_, not any crossing: a later
   source's query flipping while an earlier one still matches chooses the same
   file, and a reload would restart the loop from its first frame for nothing.
 - **The reader's pause surviving that reload.** A reload resets the element to
@@ -633,7 +651,7 @@ carries a name only, so its entries are all `undefined`. Snap and Axon take no
 declarations because they accept no names but their own.
 
 **The file must be a module** — the `export {}` above. Without an import or an
-export, `declare module` *replaces* the package's types instead of adding to
+export, `declare module` _replaces_ the package's types instead of adding to
 them, and every import from it breaks.
 
 **Declared data is sent as written, so it names what the visitor chose, never
@@ -651,7 +669,7 @@ the same reason — that sends:
 
 - with `tagIds`: `gtag("event", name, data)`, which `gtag.js` reads;
 - with `containerIds`: `dataLayer.push({ event: name, ...data })`, which is what
-  a Tag Manager *Custom Event* trigger listens for — the `dataLayer.push` a
+  a Tag Manager _Custom Event_ trigger listens for — the `dataLayer.push` a
   marketing agency's brief asks for;
 - with both, both. A container that also runs a GA4 tag on the same event then
   counts it twice; that is set up in Tag Manager, where lib cannot see it.
@@ -673,7 +691,7 @@ right properties is a faithful stand-in.
   listener, covering the matching, `matchedWithout`, the four URL decisions and
   attach/detach.
 - `tests/filters-view.test.ts` fakes plain objects with a `hidden` property,
-  pins the write-back guard by counting *writes* — so deleting the guard fails
+  pins the write-back guard by counting _writes_ — so deleting the guard fails
   rather than passing on an identical value — and pins the nesting order in
   both directions.
 - `tests/filtering.test.ts` stubs `document` with a `startViewTransition` whose

package/docs/rich-text.md

@@ -1,6 +1,6 @@
 # Rich text: the marks a message may carry
 
-`t()` prints a message. `rich()` reads the marks *in* one and hands back runs —
+`t()` prints a message. `rich()` reads the marks _in_ one and hands back runs —
 so a sentence with emphasis, a link and a line break in it stays **one message**,
 which is the only shape a translator can reorder.
 
@@ -13,7 +13,7 @@ which is the only shape a translator can reorder.
 
 ```ts
 const rich = site.rich(locale);
-rich("about.intro", { venue: "contact" })   // the destination, stated once
+rich("about.intro", { venue: "contact" }); // the destination, stated once
 ```
 
 Greek moves the link to the front and spends fewer runs on the clause. Neither is
@@ -49,7 +49,7 @@ sentence the way their tags do, and `[x]` would not.
 
 `[b]` is the one presentational-looking mark lib names outright, and it is not
 presentational — it is `<strong>`, it survives into plain text as emphasis a
-screen reader announces. Everything that is genuinely a *style* is `[v:]`.
+screen reader announces. Everything that is genuinely a _style_ is `[v:]`.
 
 ## An argument is a role or a slot, never a value
 
@@ -108,7 +108,7 @@ and fills none.
 ## Marks do not nest
 
 One mark is open at a time. `[b]bold [v:accent]and red[/v][/b]` is refused rather
-than flattened, because a run is bold *or* it carries a variant and there is
+than flattened, because a run is bold _or_ it carries a variant and there is
 nowhere for the second one to go.
 
 If the combination is wanted, it is a role: give one variant both properties and
@@ -119,7 +119,7 @@ name it once. (The shape this replaces answered the same limit by inventing a
 without closing the mark:
 
 ```ts
-parseMarks("[b]One[br]Two[/b]")
+parseMarks("[b]One[br]Two[/b]");
 // [ bold "One", break, bold "Two" ]
 ```
 
@@ -212,7 +212,7 @@ every link in every description in order to throw them all away.
 
 ```ts
 const plainText = site.plain(locale);
-plainText("about.intro", { company })      // no `venue`, no `ask`
+plainText("about.intro", { company }); // no `venue`, no `ask`
 ```
 
 Reach for `t()` when the copy is structurally plain and should stay that way — a
@@ -229,7 +229,10 @@ and escapes all the copy. A `[v:]` role and an in-page anchor come out as their
 words.
 
 ```ts
-faqPage([{ question, answer: html(rich("faq.city.a", { network: "network" })) }], at)
+faqPage(
+    [{ question, answer: html(rich("faq.city.a", { network: "network" })) }],
+    at
+);
 ```
 
 ## The renderer half
@@ -241,7 +244,11 @@ per `[v:name]` and one for links — usually in a small wrapper
 
 ```astro
 <p class="…">
-    <RichText spans={rich("about.intro", { venue: "contact" })} variants={{ accent: "…" }} link="…" />
+    <RichText
+        spans={rich("about.intro", { venue: "contact" })}
+        variants={{ accent: "…" }}
+        link="…"
+    />
 </p>
 ```
 

package/docs/share-images.md

@@ -54,4 +54,4 @@ padded — and the failure was invisible, because a card that renders is a card
 that looks like it worked. One value, no threshold to sit near, nothing to
 override. `tests/meta.test.ts` pins it at three sizes including a square one.
 
-What Atlas does *not* model is a second image for X alone: one page, one picture.
+What Atlas does _not_ model is a second image for X alone: one page, one picture.

package/docs/toolchain.md

@@ -28,7 +28,7 @@ it matters. Track
 One without the other is the trap. `experimentalFullSupportEnabled` moves
 `.astro` onto the HTML-family path, where the formatter is off unless it is
 asked for — so setting only the first drops every `.astro` file from
-`biome format` entirely. It reports them as *ignored* rather than as errors, so
+`biome format` entirely. It reports them as _ignored_ rather than as errors, so
 `check-fmt` stays green while nothing is being formatted at all.
 
 With both set, `noUnusedVariables` and `noUnusedImports` resolve template usage
@@ -56,7 +56,7 @@ single catch-all page in each example, from the route registry.
 ## Formatting
 
 4-space indent, double quotes, `es5` trailing commas, LF. Biome formats `.astro`
-frontmatter *and* markup, but not the inside of a `{ }` expression — so the
+frontmatter _and_ markup, but not the inside of a `{ }` expression — so the
 `.map()` blocks in the examples stay hand-aligned while the markup around them
 follows `indentWidth`. See the open Biome issues in the README.
 

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@escape-game-over/atlas",
-    "version": "0.1.82",
+    "version": "0.1.83",
     "type": "module",
     "description": "Typed, data-driven machinery for static multi-locale, multi-deployment Astro sites.",
     "private": false,
@@ -49,7 +49,7 @@
         "check-fmt": "biome check . --error-on-warnings"
     },
     "dependencies": {
-        "@types/google.maps": "3.66.6"
+        "@types/google.maps": "3.66.9"
     },
     "peerDependencies": {
         "@types/node": ">=22",
@@ -58,11 +58,11 @@
         "vite": ">= 8"
     },
     "devDependencies": {
-        "@biomejs/biome": "2.5.15",
-        "@types/node": "26.6.5",
-        "@vitest/coverage-istanbul": "5.0.4",
-        "astro": "7.4.0",
+        "@biomejs/biome": "2.6.4",
+        "@types/node": "26.6.6",
+        "@vitest/coverage-istanbul": "5.0.5",
+        "astro": "7.4.4",
         "typescript": "6.0.3",
-        "vitest": "5.0.4"
+        "vitest": "5.0.5"
     }
 }