@sonordev/contracts 2.2.0 → 2.4.0

package/dist/error-page.mjs

@@ -0,0 +1,416 @@
+// src/contracts/error-page.ts
+var ERROR_PAGE_CONTRACT_VERSION = 1;
+var ERROR_PAGE_MAX_WORDS = 120;
+var PREFIXED_MAX_WORDS = 400;
+var SHORT_ONLY_MAX_WORDS = 60;
+var START_STATE_MAX_WORDS = 40;
+var HEAD_CHARS = 300;
+var DESCRIBE_WINDOW = 45;
+var NAMED_ENTITIES = {
+  amp: "&",
+  quot: '"',
+  apos: "'",
+  nbsp: " ",
+  lt: "<",
+  gt: ">",
+  rsquo: "'",
+  lsquo: "'",
+  ldquo: '"',
+  ndash: "\u2013",
+  mdash: "\u2014",
+  hellip: "\u2026"
+};
+function decodeEntities(text) {
+  let out = text;
+  for (let pass = 0; pass < 2 && out.indexOf("&") !== -1; pass++) {
+    out = out.replace(/&(?:#x([0-9a-f]{1,6})|#(\d{1,7})|([a-z]{2,8}));/gi, (whole, hex, dec, name) => {
+      if (name) {
+        const key = name.toLowerCase();
+        return Object.prototype.hasOwnProperty.call(NAMED_ENTITIES, key) ? NAMED_ENTITIES[key] : whole;
+      }
+      const code = hex ? parseInt(hex, 16) : parseInt(dec, 10);
+      return code <= 1114111 ? String.fromCodePoint(code) : whole;
+    });
+  }
+  return out;
+}
+function tidy(text) {
+  return decodeEntities(text).replace(/[‘’‛ʼ]/g, "'").replace(/\s+/g, " ").trim();
+}
+function normalize(text) {
+  return tidy(text).toLowerCase();
+}
+function textOf(text) {
+  const oc = tidy(text);
+  const lc = oc.toLowerCase();
+  return { lc, oc: lc.length === oc.length ? oc : lc };
+}
+function slice(text, from, to) {
+  return { lc: text.lc.slice(from, to), oc: text.oc.slice(from, to) };
+}
+function words(text) {
+  return text.split(/\s+/).filter(Boolean).length;
+}
+function wordsAcrossElements(text) {
+  return words(text.replace(/([a-z0-9.,;:!?)"'’])([A-Z])/g, "$1 $2"));
+}
+var NOT_A_QUESTION = String.raw`(?!\s*\?|\s+(?:in|on|for|at|with|under|from|during|to)\b[^.!?]{0,50}\?)`;
+var NOT_IN_YOUR_SIZE = String.raw`(?!\s+in\s+your\s+size\b)`;
+var RECORD_NOUN = String.raw`(?:page|project|post|article|product|listing|resource|file|deployment)`;
+var CANT_BE_FOUND = String.raw`\b(?:page|url|link|resource|route|file|deployment|site|website|app|project)\b[^.!?]{0,40}?(?:could not|couldn't|can't|cannot|can not) be found`;
+var WHICH_PAGE = String.raw`(?:\s+(?:you(?:'re| are| were)? (?:looking for|requested|asked for|(?:tried|were trying|are trying) to (?:reach|visit|access|find|open))|that you (?:requested|asked for)|at\s+\S[^.!?]{0,40}|on this (?:site|server|website)))?`;
+var DOESNT_EXIST = String.raw`(?:\b(?:this|that|the|requested)\s+(?:[a-z0-9'-]+\s+){0,2}(?:page|url|link|post|article|listing|product|resource|project|profile|record|entry|story|event|course|job|item|offer|video)|\bpage)${WHICH_PAGE}\s+(?:doesn't|does not)(?: (?:seem|appear))?(?: to)? exist(?!\s+(?:to|for|so)\s)`;
+var MAINTENANCE_QUALIFIER = String.raw`(?!\s+(?:on|from|every|each|this|next|between|during|at|after|before|once|twice|monthly|weekly|nightly|daily|contract|agreement|plan|program|warranty|coverage|with\s+(?:us|our)|for\s+(?:the|a|\d)|until(?!\s+(?:\d{1,2}(?::\d\d)?\s*(?:[ap]\.?m\b|o'clock)|\d{1,2}:\d\d|noon|midnight|tomorrow|further notice)))\b)`;
+var ERROR_PAGE_SIGNATURES = [
+  // ---- Not found: the framework defaults and the common custom wordings ----
+  new RegExp(RECORD_NOUN + " (?:was |is )?not found" + NOT_A_QUESTION + NOT_IN_YOUR_SIZE),
+  // "Can't find the page you need? Use the search box" is a help line; "We can't find that page." isn't.
+  /(?:can't|cannot|can not|couldn't|could not|unable to) (?:be )?(?:find|found|locate)(?: that| this| the)? page(?!\s*\?|\s+(?:you|that|which)\b[^.!?]{0,40}\?)/,
+  // "We can't seem to find the page", "It seems we can't find what you're looking for".
+  /(?:we|i) (?:can't|cannot|couldn't|could not|were unable to|are unable to|weren't able to)(?: seem to)? (?:find|locate) (?:that page|this page|the page|that one|that(?=[.!,?]| link)|what you(?:'re| are| were)? looking for)/,
+  // "can not" is two words in the Vercel default.
+  new RegExp(CANT_BE_FOUND),
+  // Next's own not-found.tsx example: "Not FoundCould not find requested resourceReturn Home".
+  /(?:could not|couldn't) find (?:the )?requested resource/,
+  // "doesn't exist" only with a page, link or record as the subject: a short
+  // marketing page can say "if a tool doesn't exist, we'll build it".
+  new RegExp(DOESNT_EXIST),
+  /(?:doesn't|does not) exist or (?:has|have) been (?:moved|removed|deleted)/,
+  /looking for (?:doesn't|does not) exist/,
+  // A page-ish subject, like "doesn't exist" above: a history page can say "the 1890 mill no longer exists".
+  new RegExp(String.raw`\b(?:this|that|the)\s+(?:[a-z0-9'-]+\s+){0,2}(?:page|url|link|project|post|article|product|listing|resource|content|item|offer|event|record|story|profile|entry|document|course|job|video)` + WHICH_PAGE + String.raw`\s+no longer exists`),
+  /(?:404: ?not[ _]found|code: ?not_found)/,
+  // "Error 404" and "404 Error" as a label ("404 Error | Acme") or leading into what a 404 says
+  // ("Error 404Page not found"). Not "a custom 404 error page" and not a name ("Error 404 Studio").
+  /(?:^|[.!?|:]\s*)(?:http |server )?(?:error 404|404 error)(?=\s*[-–—|:.!?,)]|\s*(?:page|this|that|the|we|oops|whoops|sorry|looks|uh|hmm|nothing|there|it|oh|please)\b)|(?:error 404|404 error)(?=\s*$|\s*[-–—|:]?\s*(?:not found|this (?:page|url)|that page)\b)/,
+  // "The page you requested was not found", "The page you were looking for has been removed".
+  /page you (?:requested|(?:were|are) looking for|tried to (?:reach|visit|access))[^.!?]{0,60}(?:not found|no longer|(?:has|have|had|might have|may have|may) been|has (?:drifted|wandered|vanished|disappeared|moved|left)|isn't|is not|doesn't|does not|could not|can't|cannot|was not)/,
+  // Apache, Flask and Django: "The requested URL was not found on this server."
+  /requested (?:url|uri|resource|file|page|document)\b[^.!?]{0,80}(?:not found|not be found|cannot be found)|(?:was|is) not found on (?:this|the) server/,
+  // Netlify, Render, Flask and Express answer with the bare words "Not Found".
+  /^not found(?=$|\s*[-–—|:.!]|the requested|request id)/,
+  // WordPress core and its default themes: "It looks like nothing was found at this location.", "Nothing Found".
+  /nothing was found at this location|^nothing found(?=$|[.,:;!?|-]|sorry|it looks|we couldn|nothing|oops|the page)/,
+  // Custom 404 copy that never says "404": "Oops! This page took a day off.", "You've hit a dead end.",
+  // "This page doesn't live here anymore." The opener needs the page to be the one that's gone ("Oops! This
+  // page is a work in progress" is a holding page), and the dead end needs to be a sentence of its own.
+  /^(?:oops|whoops|uh[- ]oh)[!,.…]*\s*(?:this|that) page (?:took|went|wandered|got lost|ran away|has (?:gone|wandered|moved|left|vanished|disappeared|been (?:removed|moved|deleted))|is (?:missing|gone|lost)|isn't (?:here|available|there)|(?:doesn't|does not|can't|cannot|couldn't) (?:exist|be found|seem)|seems? (?:to be )?(?:missing|lost|gone))/,
+  /(?:^|[.!?]\s*)(?:you've|you have|you) (?:hit|reached) a dead end(?=[.!?…]|$)/,
+  /(?:page|link|url)[^.!?]{0,40}(?:doesn't|does not|no longer|isn't|is not) (?:live |exist |work )?here(?: anymore)?(?![a-z]*\s+to\b)/,
+  // Gatsby's 404 route: "NOT FOUNDYou just hit a route that doesn't exist... the sadness."
+  /hit a route that doesn't exist/,
+  // Tomcat and IIS: "HTTP Status 404 – Not Found", "HTTP Error 403.14 - Forbidden", "HTTP Error 503. The service is unavailable."
+  /http (?:error|status) [45]\d\d(?:\.\d+)?\s*[–—-]\s*(?:not found|forbidden|unauthori[sz]ed|bad request|internal|method not allowed|service unavailable)|http error [45]\d\d(?:\.\d+)?\. the /,
+  // A status number with its reason, anywhere in the head (a nav can precede it):
+  // "404 | Not Found", "502 - Bad Gateway", "403: Forbidden", "400Bad Request", "419 | Page Expired".
+  // Not "401 unauthorized means your key is missing": a status explained is documentation.
+  /(?:^|\D)(?:(?:404|410)\s*[|:–—-]?\s*(?:not found|page not found|file not found|web site not found)|408[\s:|–—-]*request time-?out|50[0-4][\s:|–—-]*(?:internal (?:server )?error|server error|bad gateway|service (?:temporarily )?unavailable|gateway time-?out)|40[13][\s:|–—-]*(?:forbidden|unauthori[sz]ed)|400[\s:|–—-]*bad request|405[\s:|–—-]*method not allowed|413[\s:|–—-]*(?:request entity|payload|content) too large|419[\s:|–—-]*page expired|429[\s:|–—-]*too many)(?!\s+(?:(?:requests?|errors?|responses?|status(?: codes?)?|codes?|messages?|pages?)\s+)?(?:means?|is|are|indicates?|occurs?|happens?|when|if|returns?|tells?|shows?|appears?|can|should|will|occurred)\b)/,
+  // Netlify's bare "Not Found - Request ID: ..." and Express's "Cannot GET /about", even with a nav in front.
+  /not found\s*[-–—]\s*request id|\bcannot (?:get|post|put|patch|delete) \/[\w\-./?=&%]*(?:$|\s)/,
+  // IIS's classic "Server Error 404 - File or directory not found." and its "had its name changed" sentence.
+  /file or directory not found|removed,? had its name changed/,
+  // This page has been removed / moved: its own sentence, not "removed from our menu".
+  /(?:^|[.!?]\s*)this page (?:has been |has |was |is )?(?:permanently )?(?:removed|deleted|moved)(?=[.!,;]|$|\s+(?:or|and)\b|\s*(?:return|go|back|click|please|visit|head)\b)/,
+  // ---- Server and client errors ----
+  // Not "Application error monitoring for Node teams".
+  /^application error(?!\s+(?:monitoring|tracking|reporting|handling|management|detection|analytics|logs?|alerts?|rates?|budgets?|codes?))/,
+  /(?:client|server)-side exception has occurred/,
+  // The hosts spell a status as a code too: "500: INTERNAL_SERVER_ERROR", "504: GATEWAY_TIMEOUT".
+  /internal[ _]server[ _]error/,
+  new RegExp(String.raw`(?:bad[ _]gateway|gateway[ _]time-?out)` + NOT_A_QUESTION),
+  // An exception's message rendered as the page: "Error: Failed to fetch data", "TypeError: x is undefined".
+  /^(?:(?:type|reference|syntax|range)error: |error: (?:failed|cannot|can't|unable|invalid|network|request|fetch|something|unexpected|no |not ))/,
+  // "Something went wrong with your insurance claim? ..." is a headline about a claim, not an error state.
+  new RegExp(String.raw`something(?: went|'s gone| has gone| had gone) wrong(?!\s+(?:with|during|on|in|at)\s+(?:your|my)\b)` + NOT_A_QUESTION),
+  /an (?:unexpected )?error (?:has )?occurred/,
+  /this (?:page|site|website) (?:isn't working|can't be reached)/,
+  // Next.js 16 error shells: "This page couldn't load. A server error occurred. Reload to try again."
+  // and the 403 / 401 pages ("This page could not be accessed.").
+  /this page couldn't load|a server error (?:has )?occurred|reload to try again(?=[.,!]|$)|this page could not be accessed/,
+  // React Router and Remix, Spring Boot, and the "unexpected error" boundaries.
+  /unexpected application error|whitelabel error page|(?:there was|encountered) an unexpected error/,
+  // Django and ASP.NET: "Server Error (500)", "Server Error in '/' Application", "Not Found (404)".
+  /server error (?:\(500\)|in '?\/)|(?:bad request|forbidden|not found) \((?:400|403|404)\)/,
+  // Next's runtime messages when a chunk or a hydration fails and the message is the page.
+  /chunkloaderror|loading chunk [\w/.-]+ failed|hydration failed because the|there was an error while hydrating/,
+  // Generated error.tsx boundaries: "Oops, we hit a snag.", "Something unexpected happened", "Unable to load page".
+  new RegExp(
+    String.raw`(?:^|[.!?]\s*)(?:(?:oops|whoops|uh[- ]oh)[!,.…]*\s*)?we hit a snag(?!\s+(?:with|in|on|during|when|while|at)\b)|(?:^|[.!?]\s*)something unexpected happened|(?:^|[.!?]\s*)(?:failed|unable) to load (?:the |this )?(?:page|content|data|resource)` + NOT_A_QUESTION + String.raw`|^error loading (?:the |this )?(?:page|content|data)`
+  ),
+  // "We're having trouble loading this page", "We couldn't load this page".
+  /(?:we're|we are|i'm) having (?:some )?(?:trouble|difficulty|difficulties|problems?) (?:loading|displaying|showing|retrieving) (?:this|that|the) (?:page|content)|we (?:couldn't|could not|can't|cannot) load (?:this|that|the) page/,
+  // WordPress core's fatal-error and database screens.
+  /there has been a critical error on this (?:website|site)|error establishing a database connection|(?:the|this) (?:site|website) is experiencing technical difficulties/,
+  // Apache's stock pages: 503, 400, 401, 408, 410 and 414 each in its own words.
+  /the server is temporarily unable to service your request|your browser sent a request that this server could not understand|this server could not verify that you are authorized|server timeout waiting for the http request|(?:was|is) no longer available on this server|request-uri too large/,
+  // ---- Unavailable and maintenance ----
+  // A qualifier ("in Hawaii", "until April") makes it content about a service, not a dead page.
+  /(?:this|the) (?:page|site|website|service) (?:is|are) (?:temporarily|currently) unavailable(?!\s+(?:in|on|for|at|during|to|until|from|outside|after|before|through)\b)/,
+  /^(?:service (?:temporarily )?unavailable|(?:temporarily|currently) unavailable(?!:\s))(?!\s+(?:in|on|for|at|during|to|until|from|outside)\b)/,
+  // The site or the page is the one under maintenance. Not "is under maintenance contract", not "down for
+  // maintenance on Fridays": a day, a time or a contract makes it content.
+  new RegExp(String.raw`(?:site|website|page|store|shop|app|portal|server|we're|we are)(?: is| are)?(?: currently| temporarily| now)*,? (?:under|down for) (?:scheduled |routine |some |emergency )?maintenance` + MAINTENANCE_QUALIFIER),
+  // "Our website is currently undergoing scheduled maintenance", "Briefly unavailable for scheduled maintenance",
+  // "We are offline for a short while". Not "Our rental fleet is undergoing routine maintenance": the subject is the site.
+  new RegExp(String.raw`(?:(?:site|website|page|store|shop|app|portal|server|we're|we are)(?: is| are)?(?: currently| temporarily| now)*|briefly|temporarily|currently|now) (?:undergoing|(?:offline|unavailable|down)(?: temporarily)? for) (?:scheduled |routine |some |emergency )?maintenance` + MAINTENANCE_QUALIFIER),
+  /(?:we're|we are|site is|website is) offline for an? (?:short|little|brief) (?:while|time|period)/,
+  /^be right back\.?$/,
+  // "We'll be back soon." ending its sentence, or leading into the next one. Not "We should be back soon with a replacement
+  // part", "back soon after our holiday break": what follows decides.
+  /(?:we'll|we will|we should|it should) (?:be )?(?:right )?back (?:online )?(?:soon|shortly)(?=[.!;]|$|,?\s*(?:our|the|this|thank|thanks|please|we)\b|\s+(?:once|after)\s+(?:a|the|our)\s+(?:(?:short|brief|quick|scheduled|planned)\s+)*(?:maintenance|upgrade|update|outage|downtime|migration|work)\b)/,
+  // An apology for the downtime makes what follows an outage notice, whatever it promises.
+  /sorry for the (?:downtime|inconvenience|interruption|outage)\W+(?:we'll|we will|we should|we are working)/,
+  // ---- Walls that stand in for the page ----
+  new RegExp(String.raw`(?:403 forbidden|access denied)(?!\s+(?:is|was|means|errors?|messages?|pages?)\b)` + NOT_A_QUESTION),
+  // A permission wall is addressed to the visitor and about THIS page or resource, and opens the text or follows its own
+  // heading ("Sorry, you are not allowed to access this page.", "ForbiddenYou are not authorized to access this resource.").
+  // Not a person or a file ("our team is not allowed to access this page of your file"), not a gallery ("you don't have
+  // permission to view the gallery"), and not a sentence further in ("Members only. You do not have permission to ...").
+  /(?:^|sorry|oops|error|forbidden|unauthori[sz]ed|access denied|40[13])\W*you(?:'re| are| aren't) (?:not |un)?(?:allowed|authori[sz]ed|permitted) to (?:access|view|see) (?:this|that) (?:page|resource|site|website|area|section|content)/,
+  /(?:^|sorry|oops|error|forbidden|unauthori[sz]ed|access denied|40[13])\W*you (?:don't|do not) have permission to (?:access|view) (?:this (?:page|resource|site|website|area|section|content)|\S+ on this server)/,
+  /^(?:forbidden|unauthori[sz]ed)(?=$|\s*[-–—|:.!]|you(?:'re| are)\b|please\b|this (?:page|resource))/,
+  /(?:checking your browser before accessing|enable javascript and cookies to continue(?!\s+\w)|verif(?:y|ying) you are (?:a )?human)/
+];
+var LEADING_STATUS = /^(?:(?:http )?error:? ?)?(?:400|401|402|403|405|408|410|413|414|419|429|500|501|502|503|504)\W{0,3}(?:(?:service|server|temporarily|internal|gateway|bad|proxy|request|too|access|page|file|resource|connection|http|status|method|payload|entity|payment|authentication)\s+){0,3}(?:error|forbidden|unauthori[sz]ed|gone|unavailable|bad gateway|time-?out|timed out|not found|not allowed|too many|too large|bad request|page expired|payment required|not implemented)/;
+var LEADING_404 = /^404(?![\d,])(?![\s.\-–—)]*\d)/;
+var ADDRESS_AFTER_404 = /^404\s+(?:[a-z0-9.'-]+\s+){0,3}(?:street|st|avenue|ave|road|rd|boulevard|blvd|drive|dr|lane|ln|way|court|ct|highway|hwy|pike|parkway|pkwy|place|pl|suite|ste|unit|apt)\b/;
+var SHORT_ONLY_SIGNATURES = [
+  // Opening the text, and not conditional: "Please refresh the page and try again." is the boundary's own line;
+  // "Clear your cache, then refresh the page and try again" is a tip, and "...try again if the calendar doesn't load" is one.
+  /^(?:please )?(?:refresh|reload)(?: the| this)? (?:page )?(?:and|or) try again(?!\s+(?:if|when|unless|whether|once|after|later|next|the next)\b)/
+];
+var SHORT_START_SIGNATURES = [new RegExp(String.raw`^(?:[a-z0-9'&-]+\s+){0,2}[a-z0-9'&-]+ not found` + NOT_A_QUESTION + NOT_IN_YOUR_SIZE)];
+var GLUED_404 = /(?:^|[\s|])404(?=[a-z])(?!th\b)/g;
+var DELIMITED_404 = /(?:^|\s)404\s*[|:–—]\s*[a-z]/g;
+var NUMBER_PREFIX = /(?:suite|ste|room|rm|unit|apt|apartment|route|rt|hwy|highway|building|bldg|floor|#)\.?\s*$/;
+var ABOUT_THE_ERROR = String.raw`(?!\W*(?:errors?\b|explained|guide|what (?:it|this|that|is|are)|how to|how do|why|means|fix(?:es|ing)?\b|causes|vs\b|\?))`;
+var PREFIX_SIGNATURES = [
+  new RegExp(String.raw`(?:page|resource|post|article|product|listing) (?:was |is )?not found`),
+  new RegExp(CANT_BE_FOUND),
+  /(?:we|i) (?:can't|cannot|couldn't|could not)(?: seem to)? (?:find|locate) (?:that page|this page|the page|what you(?:'re| are| were)? looking for)/,
+  new RegExp(DOESNT_EXIST),
+  /something went wrong|(?:client|server)-side exception|internal server error|this page couldn't load/
+].map((pattern) => new RegExp("(?:" + pattern.source + ")" + ABOUT_THE_ERROR));
+var PREFIXED_LEADING_404 = new RegExp(LEADING_404.source + ABOUT_THE_ERROR);
+var PREFIXED_LEADING_STATUS = new RegExp(LEADING_STATUS.source + ABOUT_THE_ERROR);
+var PREFIX_SEPARATOR = " \u2014 ";
+var PREFIX_MIN_CHARS = 60;
+var PREFIX_MAX_CHARS = 400;
+var SKIP_LINK = /^(?:skip to (?:main )?content|skip to main|skip navigation)\s*/;
+var PREFIX_WINDOW_CHARS = 90;
+var DESCRIBERS = String.raw`if|when|whether|unless|how to|how do|why|what is|what's|fix(?:ing|es)?|handle|handling|resolve|resolving|troubleshoot(?:ing)?|diagnos(?:e|ing)|watch(?:es|ing)? for|encounter(?:ing)?|history of`;
+var LOWERCASE_DESCRIBERS = String.raw`about|monitor(?:ing)?|alert(?:s|ing)? (?:on|for)`;
+var DESCRIBING_CONTEXT = new RegExp(
+  String.raw`(?:\b(?:(${DESCRIBERS})|(${LOWERCASE_DESCRIBERS}))(?:\s+(?:you|your|the|a|an|their|our|my|see|get|are|is|seeing|getting|receive|receiving|hit|to|every|any|each|all|\d{3})){0,3}\s*["'“]?|["“])$`
+);
+var INNER_DESCRIBER = /\b(?:if|when|whether|unless|how to)\b/;
+var SENTENCE_OPENING = /[.!?:;|—–"“(]\s*$/;
+var VISIBILITY_SUBJECT = /\b(?:you|your|we|our|they|their|customers?|clients?|people|visitors?|buyers?|patients?|business(?:es)?|brands?|compan(?:y|ies))\s+$/;
+var FORM_FIELDS = /\b(?:name|e-?mail|phone|message|subscribe|submit)\b/;
+function globalOf(patterns) {
+  return patterns.map((pattern) => new RegExp(pattern.source, "g"));
+}
+var SIGNATURES = globalOf(ERROR_PAGE_SIGNATURES);
+var SHORT_ONLY_SIGS = globalOf(SHORT_ONLY_SIGNATURES);
+var SHORT_START_SIGS = globalOf(SHORT_START_SIGNATURES);
+var PREFIX_SIGS = globalOf(PREFIX_SIGNATURES);
+var TITLE_VERB = /\b(?:fix|handle|resolve|troubleshoot|diagnose)\b/;
+function describerIsProse(orig, start, describer, context) {
+  const word = orig.slice(start, start + describer.length);
+  if (word === describer || word === describer.charAt(0).toUpperCase() + describer.slice(1)) return true;
+  return start === 0 || SENTENCE_OPENING.test(orig.slice(Math.max(0, start - 3), start)) || TITLE_VERB.test(context);
+}
+function isDescribed(head, hit) {
+  const from = Math.max(0, hit.index - DESCRIBE_WINDOW);
+  const before = head.lc.slice(from, hit.index);
+  const context = DESCRIBING_CONTEXT.exec(before);
+  if (context) {
+    const [, describer, lowercaseDescriber] = context;
+    const at = from + context.index;
+    if (describer !== void 0) {
+      if (describerIsProse(head.oc, at, describer, context[0])) return true;
+    } else if (lowercaseDescriber !== void 0) {
+      if (head.oc.slice(at, at + lowercaseDescriber.length) === lowercaseDescriber) return true;
+    } else {
+      return true;
+    }
+  }
+  const inner = INNER_DESCRIBER.exec(hit[0]);
+  if (inner && head.oc.slice(hit.index + inner.index, hit.index + inner.index + inner[0].length) === inner[0]) return true;
+  return /be found$/.test(hit[0]) && VISIBILITY_SUBJECT.test(before) || /you are (?:a )?human$/.test(hit[0]) && FORM_FIELDS.test(head.lc);
+}
+function showsSignature(head, patterns, inFirstSentence = false) {
+  return patterns.some((pattern) => {
+    pattern.lastIndex = 0;
+    for (let hit = pattern.exec(head.lc); hit; hit = pattern.exec(head.lc)) {
+      if (inFirstSentence && /[.!?]/.test(head.lc.slice(0, hit.index))) return false;
+      if (!isDescribed(head, hit)) return true;
+      pattern.lastIndex = hit.index + 1;
+    }
+    return false;
+  });
+}
+function showsLoose404(head) {
+  for (const pattern of [GLUED_404, DELIMITED_404]) {
+    pattern.lastIndex = 0;
+    for (let hit = pattern.exec(head); hit; hit = pattern.exec(head)) {
+      if (!NUMBER_PREFIX.test(head.slice(Math.max(0, hit.index - 14), hit.index))) return true;
+    }
+  }
+  return false;
+}
+function afterPrefix(text) {
+  let start = 0;
+  const cut = text.lc.indexOf(PREFIX_SEPARATOR);
+  if (cut >= PREFIX_MIN_CHARS && cut < PREFIX_MAX_CHARS) start = cut + PREFIX_SEPARATOR.length;
+  const skip = SKIP_LINK.exec(text.lc.slice(start));
+  if (skip) start += skip[0].length;
+  return start === 0 ? text : slice(text, start);
+}
+function opensWith404(head) {
+  return LEADING_404.test(head) && !ADDRESS_AFTER_404.test(head);
+}
+function showsErrorState(head, count) {
+  return opensWith404(head.lc) || LEADING_STATUS.test(head.lc) || showsSignature(head, SIGNATURES) || count <= START_STATE_MAX_WORDS && showsSignature(head, SHORT_START_SIGS) || count <= SHORT_ONLY_MAX_WORDS && (showsSignature(head, SHORT_ONLY_SIGS) || showsLoose404(head.lc));
+}
+function looksLikeErrorPage(text, wordCount) {
+  if (!text || typeof text !== "string") return false;
+  const trimmed = text.trim();
+  if (!trimmed) return false;
+  const own = wordsAcrossElements(trimmed);
+  const count = typeof wordCount === "number" && wordCount >= 0 ? Math.max(wordCount, own) : own;
+  if (count > PREFIXED_MAX_WORDS) return false;
+  const whole = textOf(trimmed);
+  const rest = afterPrefix(whole);
+  if (own > ERROR_PAGE_MAX_WORDS) {
+    const lead = rest.lc.slice(0, HEAD_CHARS);
+    return PREFIXED_LEADING_404.test(lead) && !ADDRESS_AFTER_404.test(lead) || PREFIXED_LEADING_STATUS.test(lead) || showsSignature(slice(rest, 0, PREFIX_WINDOW_CHARS), PREFIX_SIGS, true);
+  }
+  if (count > ERROR_PAGE_MAX_WORDS) return false;
+  return showsErrorState(slice(whole, 0, HEAD_CHARS), count) || // A long description pushes the page's own words past the first 300 characters.
+  rest !== whole && showsErrorState(slice(rest, 0, HEAD_CHARS), words(rest.lc));
+}
+function usablePageText(text, wordCount) {
+  if (!text || typeof text !== "string" || !text.trim()) return null;
+  return looksLikeErrorPage(text, wordCount) ? null : text;
+}
+var QUALIFIED = String.raw`(?!\s+(?:in|on|during|until|outside|after|before|from|when|while|if|with|to|through|between|over|due|at(?!\s+(?:this|the)\s+(?:time|moment))|for(?!\s+(?:this|the|that)\s+(?:[\w'’&-]+\s+){0,3}?(?:page|url|link|listing|post|article|project|menu|item|resource|insight|study|portfolio)))\b)`;
+var QUALIFIED_LIGHT = String.raw`(?!\s+(?:when|while|if|until|after|before|during|on|in|from|to|behind|outside|within)\b)`;
+var THE_PAGE = String.raw`\b(?:this|the|that)\s+(?:[\w'’&.-]+\s+){0,4}?(?:page|url|link|route|listing|post|article|project|product|resource|insight|study|portfolio|document)`;
+var ERROR_CLAIMS = [
+  // "Page Not Found", "Project Not Found", and the hyphenated compound a model coins:
+  // "a page-not-found message". Not "Page not found? Use our site map": a question is a help line.
+  new RegExp(String.raw`\b(?:page|project|post|article|listing|product)[\s-]+not[\s-]+found\b(?!\s*\?)`, "i"),
+  // "the pricing page is currently unavailable", "this URL is unavailable", "the page is currently down",
+  // "This page is no longer available", "This URL is not found"
+  new RegExp(
+    String.raw`\b(?:web\s?page|page|url|link|route)(?:\s+(?:from|at|on|for|of)\s+[\w'’&.-]+(?:\s+[\w'’&.-]+){0,4})?\s+(?:is|was|are|were|has\s+been|appears\s+to\s+be|seems\s+to\s+be|remains)\s+(?:(?:currently\s+|temporarily\s+|now\s+)?(?:unavailable|not\s+available|missing|offline|broken|gone)\b` + QUALIFIED + String.raw`|(?:currently\s+|temporarily\s+|now\s+)?(?:down|inaccessible|not\s+accessible|unreachable|not\s+found)\b` + QUALIFIED_LIGHT + String.raw`|no\s+longer\s+(?:available|accessible|reachable|live|online)\b` + QUALIFIED_LIGHT + String.raw`)`,
+    "i"
+  ),
+  // "Page Unavailable", "Page Currently Unavailable"
+  // "offline" only as a state ("the page is offline"): "save any page offline" is a feature.
+  new RegExp(String.raw`\bpage\s+(?:(?:is|was)\s+)?(?:currently\s+|temporarily\s+|now\s+)?(?:unavailable|not\s+available)\b` + QUALIFIED + String.raw`|\bpage\s+(?:is|was)\s+(?:currently\s+|temporarily\s+|now\s+)?offline\b`, "i"),
+  // "an unavailable-page presentation": the hyphenated compound again.
+  /\bunavailable-page\b/i,
+  // "could not be found" (or located, reached, loaded) with a page, URL or link as the subject
+  // The subject is THIS page ("this/the/that ... page"): "why your business page can't be found on Google" is an SEO headline.
+  /\b(?:this|the|that)\s+(?:[\w'’&.-]+\s+){0,6}?(?:page|url|link|route)\b[^.|;:]{0,60}?\b(?:could\s+not|couldn(?:'|’)t|can(?:'|’)t|cannot)\s+be\s+(?:found|located|reached|loaded|displayed|shown)\b/i,
+  // "currently returns a 404", "shows a not-found page", "displays a 404 page", "returns an HTTP 404 status"
+  /\b(?:return(?:s|ed|ing)?|show(?:s|ed|ing)?|display(?:s|ed|ing)?|serv(?:e|es|ed|ing)|lands?\s+on)\s+(?:that\s+)?(?:an?\s+|the\s+)?(?:http\s+)?(?:404|not[- ]found|error\s+page)/i,
+  /\b404[\s-]page\b/i,
+  // "details are unavailable", "menu information is unavailable"
+  new RegExp(String.raw`\b(?:details|information|info|content|menu|page)\s+(?:is|are)\s+(?:currently\s+|temporarily\s+)?unavailable\b` + QUALIFIED, "i"),
+  // A sentence that opens with the page as its subject: "This insight is unavailable."
+  // ...and the sentence's subject has to BE a page-like thing: "The patio is not available on Fridays" is about a patio.
+  new RegExp(String.raw`(?:^|[.!?]\s+)(?:this|the)\s+(?:[^.!?|\s][^.!?|]{0,69}?\s+)?(?:page|url|link|route|insight|article|post|project|study|resource|portfolio|guide|document)(?:\s+(?:you|that|which)\s+[^.!?|]{0,40}?)?\s+is\s+(?:currently\s+|temporarily\s+)?(?:unavailable|not\s+available)\b` + QUALIFIED, "i"),
+  // JSON-LD labels: "Unavailable Harbor Buffet page". A label, not "block unavailable dates on your booking page".
+  /(?:^|[|:.!?–—-]\s*)unavailable\s+(?:[\w'’&-]+\s+){0,6}?(?:page|url|menu|insight|article|project)\b/i,
+  // "This page doesn't exist.", "This page does not currently exist.", "No page exists at this address."
+  new RegExp(THE_PAGE + String.raw`\s+(?:(?:does\s+not|doesn['’]t)\s+(?:currently\s+|seem\s+to\s+|appear\s+to\s+)?exist|no\s+longer\s+exists?)\b(?!\s+(?:for|so)\s)` + QUALIFIED_LIGHT, "i"),
+  /\bno\s+(?:page|content|resource|listing)\s+exists?\s+(?:at|for|on|here)\b/i,
+  // "This page has been removed or moved.", "This listing has been removed." Moved to somewhere ("moved to our new site") is a redirect note.
+  new RegExp(THE_PAGE + String.raw`\s+(?:has|have|had)\s+(?:since\s+)?been\s+(?:permanently\s+)?(?:removed|moved|deleted|taken\s+down)\b` + QUALIFIED, "i"),
+  // "The page you're looking for isn't available.", "The page you requested has been removed."
+  new RegExp(
+    String.raw`\bpage\s+you(?:['’]re|\s+are|\s+were)?\s+(?:looking\s+for|requested|tried\s+to\s+(?:reach|visit|access))\s+(?:(?:is(?:n['’]t|\s+not)|was(?:n['’]t|\s+not))\s+(?:available|found|here|there|accessible)\b` + QUALIFIED + String.raw`|(?:has|have|might\s+have|may\s+have)\s+been\s+(?:removed|moved|deleted|renamed)\b|(?:doesn['’]t|does\s+not)\s+exist\b|no\s+longer\s+(?:exists?|available)\b)`,
+    "i"
+  ),
+  // "The page failed to load." (past tense: "if a page fails to load, refresh" is advice).
+  /\b(?:page|url|link|route)\s+(?:has\s+)?failed\s+to\s+load\b/i,
+  // "This URL contains only an error message.", "The page returned an error message instead of service details."
+  // Not "The checkout page shows an error message when payment fails."
+  new RegExp(String.raw`\b(?:page|url|link|route)\s+(?:currently\s+)?(?:(?:returned|returns?|displays?|displayed|shows?|showed|serves?|served)\s+(?:only\s+)?|(?:contains?|contained)\s+only\s+)(?:an?\s+|the\s+)?(?:error\s+(?:message|page)|http\s+error|not[- ]found\s+(?:message|error|page))\b(?=\s*(?:[.!?;:,|]|$)|\s+(?:instead|rather|in\s+place|and|but|only|with\s+no|with\s+nothing)\b)`, "i"),
+  // "404 Error | Acme Mowing", "Error 404: Acme Mowing", "HTTP 404", "404 Not Found | Acme": a label opening the copy or a segment of it.
+  /(?:^|[|:.!?–—-]\s*)(?:http\s+)?(?:error\s+404|404\s+error|http\s+404)\s*(?:[|:–—.!?-]|$)|\b404[\s:|–—-]*not\s+found\b/i,
+  // "This page isn't available.", "This page wasn't found." (the contracted forms of the above)
+  new RegExp(String.raw`\b(?:web\s?page|page|url|link|route)\s+(?:is|was)(?:n['’]t)\s+(?:currently\s+|now\s+)?(?:available|found|accessible|here|live)\b` + QUALIFIED_LIGHT, "i"),
+  // "We couldn't find this page on the Acme website.", "Sorry, we can't find that page.", "We could not locate the page you requested."
+  // Not "Can't find the page you need? Use search": a question in the same sentence is a help line.
+  /\b(?:we|i)\s+(?:could\s+not|couldn['’]t|can['’]t|cannot|(?:were|was|are|am)\s+unable\s+to|weren['’]t\s+able\s+to)\s+(?:seem\s+to\s+)?(?:find|locate)\s+(?:this|that|the)\s+(?:requested\s+)?(?:web\s?page|page|url|link)\b(?![^.!?]*\?)/i,
+  // "Unable to find the requested page."
+  /\bunable\s+to\s+(?:find|locate)\s+(?:this|that|the)\s+requested\s+(?:web\s?page|page|url|link|resource)\b/i,
+  // "The requested URL was not found on this server."
+  /\brequested\s+(?:url|uri|resource|page)\b[^.!?]{0,60}?\b(?:was\s+not|wasn['’]t|could\s+not\s+be|cannot\s+be)\s+found\b/i,
+  // "Page missing", "Error loading page": labels (a segment or the whole copy), not "pages missing from the index".
+  /(?:^|[|:.!?–—-]\s*)(?:page\s+(?:is\s+)?missing|error\s+loading\s+(?:the\s+)?page|not\s+found)\s*(?:[|:–—.!?-]|$)/i
+];
+function ownTextOf(pageText) {
+  return pageText && typeof pageText === "string" && !looksLikeErrorPage(pageText) ? normalize(pageText) : "";
+}
+var CLAIM_MAX_CHARS = 5e3;
+function claimIn(copy, own) {
+  const text = copy.replace(/\s+/g, " ").trim().slice(0, CLAIM_MAX_CHARS);
+  if (!text) return null;
+  for (const claim of ERROR_CLAIMS) {
+    const hit = text.match(claim);
+    if (!hit) continue;
+    const phrase = hit[0].replace(/^[.!?|:–—-]\s*/, "").trim();
+    if (own && own.includes(normalize(phrase))) continue;
+    return phrase;
+  }
+  return null;
+}
+function findErrorPageClaim(copy, pageText) {
+  if (!copy || typeof copy !== "string") return null;
+  return claimIn(copy, ownTextOf(pageText));
+}
+function describesErrorPage(copy, pageText) {
+  return findErrorPageClaim(copy, pageText) !== null;
+}
+var COPY_MAX_DEPTH = 8;
+var COPY_MAX_LEAVES = 2e3;
+var COPY_MAX_TOTAL_CHARS = 1e5;
+function findPageCopyClaim(fields, pageText) {
+  if (!fields || typeof fields !== "object") return null;
+  const own = ownTextOf(pageText);
+  let leaves = 0;
+  let containers = 0;
+  let chars = 0;
+  const spent = () => leaves >= COPY_MAX_LEAVES || containers > COPY_MAX_LEAVES || chars > COPY_MAX_TOTAL_CHARS;
+  const walk = (value, level) => {
+    if (typeof value === "string") {
+      leaves++;
+      chars += Math.min(value.length, CLAIM_MAX_CHARS);
+      return spent() && chars > COPY_MAX_TOTAL_CHARS ? null : claimIn(value, own);
+    }
+    if (!value || typeof value !== "object") return null;
+    if (level >= COPY_MAX_DEPTH || ++containers > COPY_MAX_LEAVES) return null;
+    const record2 = value;
+    for (const key in record2) {
+      if (spent()) break;
+      if (!Object.prototype.hasOwnProperty.call(record2, key)) continue;
+      const phrase = walk(record2[key], level + 1);
+      if (phrase) return phrase;
+    }
+    return null;
+  };
+  const record = fields;
+  for (const key of Object.keys(record)) {
+    if (spent()) break;
+    const phrase = walk(record[key], 1);
+    if (phrase) return { field: key, phrase };
+  }
+  return null;
+}
+
+export { ERROR_PAGE_CONTRACT_VERSION, ERROR_PAGE_MAX_WORDS, describesErrorPage, findErrorPageClaim, findPageCopyClaim, looksLikeErrorPage, usablePageText };

package/dist/llms.d.ts

@@ -28,6 +28,7 @@ export interface LLMsPayloadMeta {
     /** Optional single-line, public-safe note in the llms.txt blockquote (e.g. scope / not legal advice). */
     llms_disclaimer?: string | null;
 }
+export declare function clipAtWordBoundary(text: string, max: number): string;
 /**
  * Public-safe summary for llms.txt `[title](url): notes` — never CRM-only or internal copy.
  */

package/dist/llms.js

@@ -18,14 +18,64 @@ var MANAGED_LLM_SCHEMA_KNOWN_KEYS = /* @__PURE__ */ new Set([
   "speakable",
   "additionalType"
 ]);
+var SENTENCE_SHARE = 0.4;
+var ELLIPSIS = "\u2026";
+var CLOSERS = /* @__PURE__ */ new Set(['"', "'", ")", "]", "\u201D", "\u2019", "\xBB"]);
+var ABBREVIATIONS = /* @__PURE__ */ new Set([
+  "mr",
+  "mrs",
+  "ms",
+  "dr",
+  "prof",
+  "st",
+  "mt",
+  "ft",
+  "jr",
+  "sr",
+  "inc",
+  "ltd",
+  "co",
+  "corp",
+  "vs",
+  "no"
+]);
+var TRAILING_SEPARATORS = /[\s.,;:\-\u2013\u2014/&([{]+$/;
+function clipAtWordBoundary(text, max) {
+  const flat = text.replace(/\s+/g, " ").trim();
+  if (flat.length <= max) return flat;
+  if (max < 2) return flat.slice(0, Math.max(0, max));
+  const floor = Math.floor(max * (1 - SENTENCE_SHARE));
+  for (let end = max; end > floor; end--) {
+    if (endsSentence(flat, end)) return flat.slice(0, end);
+  }
+  const room = max - 1;
+  const space = flat.lastIndexOf(" ", room);
+  if (space <= 0) return `${flat.slice(0, room)}${ELLIPSIS}`;
+  if (endsSentence(flat, space)) return flat.slice(0, space);
+  const head = flat.slice(0, space).replace(TRAILING_SEPARATORS, "");
+  return `${head || flat.slice(0, room)}${ELLIPSIS}`;
+}
+function endsSentence(flat, end) {
+  if (flat[end] !== " ") return false;
+  let i = end - 1;
+  while (i > 0 && CLOSERS.has(flat[i])) i--;
+  const terminator = flat[i];
+  if (terminator === "!" || terminator === "?") return true;
+  return terminator === "." && !isAbbreviation(flat, i);
+}
+function isAbbreviation(flat, dot) {
+  let start = dot;
+  while (start > 0 && flat[start - 1] !== " ") start--;
+  const token = flat.slice(start, dot).replace(/^[("'[\u201c\u2018]+/, "").toLowerCase();
+  if (token.includes(".") || /^[a-z]$/.test(token)) return true;
+  return ABBREVIATIONS.has(token);
+}
 function sanitizeLlmsPublicSummary(raw) {
   if (raw === null || raw === void 0) return null;
   if (typeof raw !== "string") return null;
   let s = raw.trim().replace(/\s+/g, " ");
   s = s.replace(/[\[\]\r\n]/g, " ").replace(/\s+/g, " ").trim();
-  if (s.length > LLMS_PUBLIC_SUMMARY_MAX_LENGTH) {
-    s = `${s.slice(0, LLMS_PUBLIC_SUMMARY_MAX_LENGTH - 1).trimEnd()}\u2026`;
-  }
+  s = clipAtWordBoundary(s, LLMS_PUBLIC_SUMMARY_MAX_LENGTH);
   if (/\b[\w.%+-]+@[\w.-]+\.[A-Za-z]{2,}\b/.test(s)) return null;
   if (/\b\+?\d[\d\s().-]{8,}\b/.test(s)) return null;
   if (s.length < 8) return null;
@@ -36,9 +86,7 @@ function sanitizeLlmsDisclaimerLine(raw) {
   if (typeof raw !== "string") return null;
   let s = raw.trim().replace(/\s+/g, " ");
   s = s.replace(/[\[\]\r\n]/g, " ").replace(/\s+/g, " ").trim();
-  if (s.length > LLMS_DISCLAIMER_MAX_LENGTH) {
-    s = `${s.slice(0, LLMS_DISCLAIMER_MAX_LENGTH - 1).trimEnd()}\u2026`;
-  }
+  s = clipAtWordBoundary(s, LLMS_DISCLAIMER_MAX_LENGTH);
   if (/\b[\w.%+-]+@[\w.-]+\.[A-Za-z]{2,}\b/.test(s)) return null;
   if (/\b\+?\d[\d\s().-]{8,}\b/.test(s)) return null;
   if (s.length < 3) return null;
@@ -100,6 +148,7 @@ exports.LLMS_DISCLAIMER_MAX_LENGTH = LLMS_DISCLAIMER_MAX_LENGTH;
 exports.LLMS_PUBLIC_SUMMARY_MAX_LENGTH = LLMS_PUBLIC_SUMMARY_MAX_LENGTH;
 exports.LLM_GEO_CONTRACT_VERSION = LLM_GEO_CONTRACT_VERSION;
 exports.MANAGED_LLM_SCHEMA_KNOWN_KEYS = MANAGED_LLM_SCHEMA_KNOWN_KEYS;
+exports.clipAtWordBoundary = clipAtWordBoundary;
 exports.isEmittableReviewNode = isEmittableReviewNode;
 exports.pickManagedLlmSchemaForJsonLd = pickManagedLlmSchemaForJsonLd;
 exports.sanitizeLlmsDisclaimerLine = sanitizeLlmsDisclaimerLine;

package/dist/llms.mjs

@@ -16,14 +16,64 @@ var MANAGED_LLM_SCHEMA_KNOWN_KEYS = /* @__PURE__ */ new Set([
   "speakable",
   "additionalType"
 ]);
+var SENTENCE_SHARE = 0.4;
+var ELLIPSIS = "\u2026";
+var CLOSERS = /* @__PURE__ */ new Set(['"', "'", ")", "]", "\u201D", "\u2019", "\xBB"]);
+var ABBREVIATIONS = /* @__PURE__ */ new Set([
+  "mr",
+  "mrs",
+  "ms",
+  "dr",
+  "prof",
+  "st",
+  "mt",
+  "ft",
+  "jr",
+  "sr",
+  "inc",
+  "ltd",
+  "co",
+  "corp",
+  "vs",
+  "no"
+]);
+var TRAILING_SEPARATORS = /[\s.,;:\-\u2013\u2014/&([{]+$/;
+function clipAtWordBoundary(text, max) {
+  const flat = text.replace(/\s+/g, " ").trim();
+  if (flat.length <= max) return flat;
+  if (max < 2) return flat.slice(0, Math.max(0, max));
+  const floor = Math.floor(max * (1 - SENTENCE_SHARE));
+  for (let end = max; end > floor; end--) {
+    if (endsSentence(flat, end)) return flat.slice(0, end);
+  }
+  const room = max - 1;
+  const space = flat.lastIndexOf(" ", room);
+  if (space <= 0) return `${flat.slice(0, room)}${ELLIPSIS}`;
+  if (endsSentence(flat, space)) return flat.slice(0, space);
+  const head = flat.slice(0, space).replace(TRAILING_SEPARATORS, "");
+  return `${head || flat.slice(0, room)}${ELLIPSIS}`;
+}
+function endsSentence(flat, end) {
+  if (flat[end] !== " ") return false;
+  let i = end - 1;
+  while (i > 0 && CLOSERS.has(flat[i])) i--;
+  const terminator = flat[i];
+  if (terminator === "!" || terminator === "?") return true;
+  return terminator === "." && !isAbbreviation(flat, i);
+}
+function isAbbreviation(flat, dot) {
+  let start = dot;
+  while (start > 0 && flat[start - 1] !== " ") start--;
+  const token = flat.slice(start, dot).replace(/^[("'[\u201c\u2018]+/, "").toLowerCase();
+  if (token.includes(".") || /^[a-z]$/.test(token)) return true;
+  return ABBREVIATIONS.has(token);
+}
 function sanitizeLlmsPublicSummary(raw) {
   if (raw === null || raw === void 0) return null;
   if (typeof raw !== "string") return null;
   let s = raw.trim().replace(/\s+/g, " ");
   s = s.replace(/[\[\]\r\n]/g, " ").replace(/\s+/g, " ").trim();
-  if (s.length > LLMS_PUBLIC_SUMMARY_MAX_LENGTH) {
-    s = `${s.slice(0, LLMS_PUBLIC_SUMMARY_MAX_LENGTH - 1).trimEnd()}\u2026`;
-  }
+  s = clipAtWordBoundary(s, LLMS_PUBLIC_SUMMARY_MAX_LENGTH);
   if (/\b[\w.%+-]+@[\w.-]+\.[A-Za-z]{2,}\b/.test(s)) return null;
   if (/\b\+?\d[\d\s().-]{8,}\b/.test(s)) return null;
   if (s.length < 8) return null;
@@ -34,9 +84,7 @@ function sanitizeLlmsDisclaimerLine(raw) {
   if (typeof raw !== "string") return null;
   let s = raw.trim().replace(/\s+/g, " ");
   s = s.replace(/[\[\]\r\n]/g, " ").replace(/\s+/g, " ").trim();
-  if (s.length > LLMS_DISCLAIMER_MAX_LENGTH) {
-    s = `${s.slice(0, LLMS_DISCLAIMER_MAX_LENGTH - 1).trimEnd()}\u2026`;
-  }
+  s = clipAtWordBoundary(s, LLMS_DISCLAIMER_MAX_LENGTH);
   if (/\b[\w.%+-]+@[\w.-]+\.[A-Za-z]{2,}\b/.test(s)) return null;
   if (/\b\+?\d[\d\s().-]{8,}\b/.test(s)) return null;
   if (s.length < 3) return null;
@@ -94,4 +142,4 @@ function pickManagedLlmSchemaForJsonLd(schema) {
   return Object.keys(out).length ? out : null;
 }
 
-export { LLMS_DISCLAIMER_MAX_LENGTH, LLMS_PUBLIC_SUMMARY_MAX_LENGTH, LLM_GEO_CONTRACT_VERSION, MANAGED_LLM_SCHEMA_KNOWN_KEYS, isEmittableReviewNode, pickManagedLlmSchemaForJsonLd, sanitizeLlmsDisclaimerLine, sanitizeLlmsPublicSummary, sanitizePrimaryLanguageTag };
+export { LLMS_DISCLAIMER_MAX_LENGTH, LLMS_PUBLIC_SUMMARY_MAX_LENGTH, LLM_GEO_CONTRACT_VERSION, MANAGED_LLM_SCHEMA_KNOWN_KEYS, clipAtWordBoundary, isEmittableReviewNode, pickManagedLlmSchemaForJsonLd, sanitizeLlmsDisclaimerLine, sanitizeLlmsPublicSummary, sanitizePrimaryLanguageTag };