@kavchat/core 0.0.1 → 0.0.2

package/dist/index.d.mts

@@ -80,11 +80,16 @@ type ClientRelease = {
  * - אותיות קטנות, הסרת ניקוד וטעמים (עברית), תשכיל (ערבית) וסימני הטעמה לטיניים (é → e)
  * - אותיות סופיות בעברית (ם → מ), צורות אליף בערבית (أ → ا), ё → е, ß → ss
  * - מילות קישור לכל שפה, והסרת תחיליות בעברית (ו, ה, ב...) ובערבית (ال, و, ب...)
+ * - תאית ושאר הכתבים בלי רווחים בין מילים (לאו, חמרית, בורמזית, יפנית וסינית): ראו isSpacelessScript
  *
  * הפורטים ל-Dart, Swift ו-Kotlin חייבים לעבור את test-vectors/normalize.json
  */
-declare const SUPPORTED_LOCALES: readonly ["he", "en", "ar", "ru", "es", "fr", "de", "pt", "it", "tr"];
+declare const SUPPORTED_LOCALES: readonly ["he", "en", "ar", "ru", "es", "fr", "de", "pt", "it", "tr", "th"];
 type Locale = (typeof SUPPORTED_LOCALES)[number];
+/**
+ * מילה שיש בה תו מכתב בלי רווחים. מילות מפתח כאלה מותאמות כתת-מחרוזת של השאלה, ולא כמילה שלמה (ראו match.ts)
+ */
+declare function isSpacelessScript(word: string): boolean;
 /** מנקה ומחזיר רשימת מילים */
 declare function normalizeWords(text: string): string[];
 /**
@@ -110,6 +115,13 @@ declare const toBubbles: (text: string) => string[];
  * התאמה של שאלה חופשית לשלב בתסריט
  * ציון: מילה מובילה = 3, שם התסריט או כפתור שמוביל לשלב = 2, שגיאת כתיב של אות אחת = 0.5.
  * שלב אחד עם ציון כפול מהבא = single. כמה קרובים = multi (עד 3). אין כלום = none
+ *
+ * כתבים בלי רווחים (תאית, לאו, חמרית, בורמזית, יפנית, סינית): "ใบเสร็จอยู่ที่ไหน" היא מילה אחת בעיני normalizeWords,
+ * ולכן השוואת מילים שלמות לא תמצא בה את "ใบเสร็จ". מילת מפתח, שם תסריט או כפתור שיש בהם תו מכתב כזה
+ * מותאמים כתת-מחרוזת של כל השאלה המנורמלת, אחרי חיבור המילים בלי רווחים (משתמשים מוסיפים רווחים באמצע).
+ * המשקלות זהים (3 ו-2). בתוך שלב, מונח שמוכל במונח ארוך יותר שנמצא, עם משקל שווה או גבוה, לא נספר שוב.
+ * אין סובלנות לשגיאות כתיב בכתבים האלה: השאלה היא רצף אחד בלי גבולות מילים, ושינוי של סימן אחד
+ * (למשל טון: ข้าว ו-ข่าว) הוא מילה אחרת. שגיאה של אות אחת בתת-מחרוזת קצרה הייתה מתאימה כמעט לכל שאלה
  */
 type MatchReason = {
     kind: 'keyword' | 'title' | 'typo';
@@ -307,5 +319,5 @@ declare function eventsPayload(state: BotState, sessionId: string, releaseVersio
     }[];
 };
 
-export { AI_ANSWER_MAX, AI_QUESTION_MAX, RTL_LOCALES, SUPPORTED_LOCALES, UI_TEXTS, aiQuestion, answerWithAi, configureEngine, eventsPayload, format, groupKey, isRtl, keyTerms, matchQuestion, normalizeWords, pendingQuestion, pressChip, sendText, startConversation, stripPrefixes, toBubbles, transcriptOf, uiTexts, withinOneEdit };
+export { AI_ANSWER_MAX, AI_QUESTION_MAX, RTL_LOCALES, SUPPORTED_LOCALES, UI_TEXTS, aiQuestion, answerWithAi, configureEngine, eventsPayload, format, groupKey, isRtl, isSpacelessScript, keyTerms, matchQuestion, normalizeWords, pendingQuestion, pressChip, sendText, startConversation, stripPrefixes, toBubbles, transcriptOf, uiTexts, withinOneEdit };
 export type { BotAction, BotButton, BotChip, BotEffect, BotMatch, BotMatchResult, BotMessage, BotOutcome, BotPathEntry, BotPosition, BotResult, BotScript, BotScriptContent, BotSettings, BotState, BotStep, ChipKind, ClientRelease, EngineOptions, Locale, MatchReason, TranscriptEntry, UiKey, UiTexts };

package/dist/index.d.ts

@@ -80,11 +80,16 @@ type ClientRelease = {
  * - אותיות קטנות, הסרת ניקוד וטעמים (עברית), תשכיל (ערבית) וסימני הטעמה לטיניים (é → e)
  * - אותיות סופיות בעברית (ם → מ), צורות אליף בערבית (أ → ا), ё → е, ß → ss
  * - מילות קישור לכל שפה, והסרת תחיליות בעברית (ו, ה, ב...) ובערבית (ال, و, ب...)
+ * - תאית ושאר הכתבים בלי רווחים בין מילים (לאו, חמרית, בורמזית, יפנית וסינית): ראו isSpacelessScript
  *
  * הפורטים ל-Dart, Swift ו-Kotlin חייבים לעבור את test-vectors/normalize.json
  */
-declare const SUPPORTED_LOCALES: readonly ["he", "en", "ar", "ru", "es", "fr", "de", "pt", "it", "tr"];
+declare const SUPPORTED_LOCALES: readonly ["he", "en", "ar", "ru", "es", "fr", "de", "pt", "it", "tr", "th"];
 type Locale = (typeof SUPPORTED_LOCALES)[number];
+/**
+ * מילה שיש בה תו מכתב בלי רווחים. מילות מפתח כאלה מותאמות כתת-מחרוזת של השאלה, ולא כמילה שלמה (ראו match.ts)
+ */
+declare function isSpacelessScript(word: string): boolean;
 /** מנקה ומחזיר רשימת מילים */
 declare function normalizeWords(text: string): string[];
 /**
@@ -110,6 +115,13 @@ declare const toBubbles: (text: string) => string[];
  * התאמה של שאלה חופשית לשלב בתסריט
  * ציון: מילה מובילה = 3, שם התסריט או כפתור שמוביל לשלב = 2, שגיאת כתיב של אות אחת = 0.5.
  * שלב אחד עם ציון כפול מהבא = single. כמה קרובים = multi (עד 3). אין כלום = none
+ *
+ * כתבים בלי רווחים (תאית, לאו, חמרית, בורמזית, יפנית, סינית): "ใบเสร็จอยู่ที่ไหน" היא מילה אחת בעיני normalizeWords,
+ * ולכן השוואת מילים שלמות לא תמצא בה את "ใบเสร็จ". מילת מפתח, שם תסריט או כפתור שיש בהם תו מכתב כזה
+ * מותאמים כתת-מחרוזת של כל השאלה המנורמלת, אחרי חיבור המילים בלי רווחים (משתמשים מוסיפים רווחים באמצע).
+ * המשקלות זהים (3 ו-2). בתוך שלב, מונח שמוכל במונח ארוך יותר שנמצא, עם משקל שווה או גבוה, לא נספר שוב.
+ * אין סובלנות לשגיאות כתיב בכתבים האלה: השאלה היא רצף אחד בלי גבולות מילים, ושינוי של סימן אחד
+ * (למשל טון: ข้าว ו-ข่าว) הוא מילה אחרת. שגיאה של אות אחת בתת-מחרוזת קצרה הייתה מתאימה כמעט לכל שאלה
  */
 type MatchReason = {
     kind: 'keyword' | 'title' | 'typo';
@@ -307,5 +319,5 @@ declare function eventsPayload(state: BotState, sessionId: string, releaseVersio
     }[];
 };
 
-export { AI_ANSWER_MAX, AI_QUESTION_MAX, RTL_LOCALES, SUPPORTED_LOCALES, UI_TEXTS, aiQuestion, answerWithAi, configureEngine, eventsPayload, format, groupKey, isRtl, keyTerms, matchQuestion, normalizeWords, pendingQuestion, pressChip, sendText, startConversation, stripPrefixes, toBubbles, transcriptOf, uiTexts, withinOneEdit };
+export { AI_ANSWER_MAX, AI_QUESTION_MAX, RTL_LOCALES, SUPPORTED_LOCALES, UI_TEXTS, aiQuestion, answerWithAi, configureEngine, eventsPayload, format, groupKey, isRtl, isSpacelessScript, keyTerms, matchQuestion, normalizeWords, pendingQuestion, pressChip, sendText, startConversation, stripPrefixes, toBubbles, transcriptOf, uiTexts, withinOneEdit };
 export type { BotAction, BotButton, BotChip, BotEffect, BotMatch, BotMatchResult, BotMessage, BotOutcome, BotPathEntry, BotPosition, BotResult, BotScript, BotScriptContent, BotSettings, BotState, BotStep, ChipKind, ClientRelease, EngineOptions, Locale, MatchReason, TranscriptEntry, UiKey, UiTexts };

package/dist/index.js

@@ -32,6 +32,7 @@ __export(index_exports, {
   format: () => format,
   groupKey: () => groupKey,
   isRtl: () => isRtl,
+  isSpacelessScript: () => isSpacelessScript,
   keyTerms: () => keyTerms,
   matchQuestion: () => matchQuestion,
   normalizeWords: () => normalizeWords,
@@ -48,7 +49,7 @@ __export(index_exports, {
 module.exports = __toCommonJS(index_exports);
 
 // src/normalize.ts
-var SUPPORTED_LOCALES = ["he", "en", "ar", "ru", "es", "fr", "de", "pt", "it", "tr"];
+var SUPPORTED_LOCALES = ["he", "en", "ar", "ru", "es", "fr", "de", "pt", "it", "tr", "th"];
 var FOLD = {
   // עברית: אותיות סופיות
   \u05DA: "\u05DB",
@@ -74,6 +75,13 @@ var FOLD = {
 function isCombiningMark(code) {
   return code >= 768 && code <= 879 || code >= 1425 && code <= 1479 || code >= 1552 && code <= 1562 || code >= 1611 && code <= 1631 || code === 1648 || code >= 1750 && code <= 1773;
 }
+function isSpacelessLetter(code) {
+  return code >= 3585 && code <= 3630 || code >= 3632 && code <= 3642 || code >= 3648 && code <= 3653 || code >= 3655 && code <= 3662 || code >= 3713 && code <= 3758 || code >= 3760 && code <= 3781 || code >= 3784 && code <= 3790 || code >= 3804 && code <= 3807 || code >= 4096 && code <= 4169 || code >= 4176 && code <= 4253 || code >= 6016 && code <= 6099 || code === 6108 || code === 6109 || code === 12293 || code >= 12353 && code <= 12438 || code >= 12441 && code <= 12442 || code >= 12445 && code <= 12447 || code >= 12449 && code <= 12538 || code >= 12540 && code <= 12543 || code >= 13312 && code <= 19903 || code >= 19968 && code <= 40959;
+}
+function isSpacelessScript(word) {
+  for (let i = 0; i < word.length; i++) if (isSpacelessLetter(word.charCodeAt(i))) return true;
+  return false;
+}
 function isWordChar(ch) {
   const code = ch.charCodeAt(0);
   if (code >= 48 && code <= 57) return true;
@@ -81,9 +89,75 @@ function isWordChar(ch) {
   if (code >= 1488 && code <= 1514) return true;
   if (code >= 1568 && code <= 1610) return true;
   if (code >= 1649 && code <= 1747) return true;
-  if (code >= 19968 && code <= 40959) return true;
+  if (isSpacelessLetter(code)) return true;
   return ch.toLowerCase() !== ch.toUpperCase();
 }
+var isZeroWidth = (code) => code >= 8203 && code <= 8205 || code === 8288 || code === 65279;
+var DIGIT_ZEROS = [3664, 3792, 4160, 6112];
+function asciiDigit(code) {
+  for (const zero of DIGIT_ZEROS) if (code >= zero && code <= zero + 9) return String.fromCharCode(48 + code - zero);
+  return void 0;
+}
+var MARK_ORDER = {
+  3640: 103,
+  3641: 103,
+  3642: 9,
+  3656: 107,
+  3657: 107,
+  3658: 107,
+  3659: 107,
+  3768: 118,
+  3769: 118,
+  3770: 9,
+  3784: 122,
+  3785: 122,
+  3786: 122,
+  3787: 122,
+  4151: 7,
+  4153: 9,
+  4154: 9,
+  4237: 220,
+  6098: 9,
+  6109: 230,
+  12441: 8,
+  12442: 8
+};
+var isThaiTone = (code) => code >= 3656 && code <= 3659;
+function orderMarks(text) {
+  if (!/[\u0e33\u0e38-\u0e3a\u0e48-\u0e4b\u0e4d\u0eb8-\u0eba\u0ec8-\u0ecb\u1037\u1039\u103a\u108d\u17d2\u17dd\u3099\u309a]/.test(text)) return text;
+  const units = [];
+  for (let i = 0; i < text.length; i++) units.push(text.charCodeAt(i));
+  for (let i = 0; i < units.length; i++) {
+    if (!MARK_ORDER[units[i]]) continue;
+    let end = i;
+    while (end < units.length && MARK_ORDER[units[end]]) end++;
+    for (let a = i + 1; a < end; a++) {
+      const v = units[a];
+      let b = a - 1;
+      while (b >= i && MARK_ORDER[units[b]] > MARK_ORDER[v]) {
+        units[b + 1] = units[b];
+        b--;
+      }
+      units[b + 1] = v;
+    }
+    i = end - 1;
+  }
+  let out = "";
+  for (let i = 0; i < units.length; i++) {
+    const c = units[i];
+    if (c === 3661 && units[i + 1] === 3634) {
+      out += "\u0E33";
+      i++;
+    } else if (c === 3661 && isThaiTone(units[i + 1]) && units[i + 2] === 3634) {
+      out += String.fromCharCode(units[i + 1], 3635);
+      i += 2;
+    } else if (c === 3635 && isThaiTone(units[i + 1])) {
+      out += String.fromCharCode(units[i + 1], 3635);
+      i++;
+    } else out += String.fromCharCode(c);
+  }
+  return out;
+}
 var decompose = (s) => {
   try {
     return s.normalize("NFD");
@@ -92,16 +166,16 @@ var decompose = (s) => {
   }
 };
 function normalizeWords(text) {
-  var _a;
+  var _a, _b;
   const source = decompose(String(text != null ? text : "").toLowerCase());
   let out = "";
   for (const raw of source) {
     const code = raw.charCodeAt(0);
-    if (isCombiningMark(code)) continue;
-    const ch = (_a = FOLD[raw]) != null ? _a : raw;
+    if (isCombiningMark(code) || isZeroWidth(code)) continue;
+    const ch = (_b = (_a = FOLD[raw]) != null ? _a : asciiDigit(code)) != null ? _b : raw;
     out += ch.length === 1 && !isWordChar(ch) ? " " : ch;
   }
-  const cleaned = out.replace(/\s+/g, " ").trim();
+  const cleaned = orderMarks(out).replace(/\s+/g, " ").trim();
   return cleaned ? cleaned.split(" ") : [];
 }
 var STOP_WORDS = {
@@ -114,7 +188,9 @@ var STOP_WORDS = {
   de: ["wie", "was", "ich", "mein", "meine", "mir", "der", "die", "das", "ein", "eine", "und", "oder", "in", "mit", "fur", "von", "zu", "ist", "kann", "mochte", "wann", "wo", "warum", "nicht", "es", "den", "dem"],
   pt: ["como", "que", "eu", "meu", "minha", "o", "a", "os", "as", "um", "uma", "de", "do", "da", "em", "no", "na", "com", "para", "por", "e", "ou", "posso", "quero", "quando", "onde", "porque", "nao"],
   it: ["come", "che", "cosa", "io", "mio", "mia", "il", "lo", "la", "i", "gli", "le", "un", "una", "di", "del", "della", "in", "con", "per", "e", "o", "posso", "voglio", "quando", "dove", "perche", "non", "l"],
-  tr: ["nasil", "ne", "ben", "benim", "bir", "ve", "veya", "ile", "icin", "bu", "su", "mi", "mu", "nerede", "ne zaman", "neden", "kac", "istiyorum", "var", "yok", "de", "da"]
+  tr: ["nasil", "ne", "ben", "benim", "bir", "ve", "veya", "ile", "icin", "bu", "su", "mi", "mu", "nerede", "ne zaman", "neden", "kac", "istiyorum", "var", "yok", "de", "da"],
+  // בתאית מילות קישור נמחקות רק כשהן מילה נפרדת (מוקפות ברווחים). רוב השאלות בלי רווחים, וההתאמה שם לפי תת-מחרוזת
+  th: ["\u0E09\u0E31\u0E19", "\u0E1C\u0E21", "\u0E14\u0E34\u0E09\u0E31\u0E19", "\u0E04\u0E38\u0E13", "\u0E04\u0E23\u0E31\u0E1A", "\u0E04\u0E48\u0E30", "\u0E04\u0E30", "\u0E19\u0E30", "\u0E44\u0E2B\u0E21", "\u0E21\u0E31\u0E49\u0E22", "\u0E2D\u0E30\u0E44\u0E23", "\u0E2D\u0E22\u0E48\u0E32\u0E07\u0E44\u0E23", "\u0E22\u0E31\u0E07\u0E44\u0E07", "\u0E17\u0E35\u0E48", "\u0E41\u0E25\u0E30", "\u0E2B\u0E23\u0E37\u0E2D", "\u0E02\u0E2D\u0E07", "\u0E44\u0E14\u0E49", "\u0E08\u0E30", "\u0E43\u0E19", "\u0E01\u0E31\u0E1A", "\u0E40\u0E1B\u0E47\u0E19", "\u0E21\u0E35", "\u0E44\u0E21\u0E48", "\u0E43\u0E2B\u0E49", "\u0E27\u0E48\u0E32", "\u0E2D\u0E22\u0E32\u0E01", "\u0E15\u0E49\u0E2D\u0E07\u0E01\u0E32\u0E23", "\u0E40\u0E21\u0E37\u0E48\u0E2D\u0E44\u0E2B\u0E23\u0E48", "\u0E17\u0E35\u0E48\u0E44\u0E2B\u0E19", "\u0E17\u0E33\u0E44\u0E21", "\u0E40\u0E17\u0E48\u0E32\u0E44\u0E2B\u0E23\u0E48"]
 };
 var NORMALIZED_STOP_WORDS = {};
 for (const locale of SUPPORTED_LOCALES) {
@@ -189,6 +265,8 @@ function matchQuestion(scripts, text, locale = "en") {
   const terms = Array.from(new Set(keyTerms(text, locale)));
   if (terms.length === 0) return { kind: "none", matches: [] };
   const terms_ = (values) => values.flatMap((v) => keyTerms(v, locale));
+  const wordTerms = terms.filter((t) => !isSpacelessScript(t));
+  const haystack = normalizeWords(text).join("");
   const candidates = [];
   for (const script of scripts) {
     if (script.kind === "opening") continue;
@@ -206,20 +284,25 @@ function matchQuestion(scripts, text, locale = "en") {
       const isRoot = step.id === content.rootStepId;
       const strong = terms_([...step.keywords, ...isRoot ? content.keywords : []]);
       const weak = terms_([...isRoot ? [content.name] : [], ...(_b = labelsTo.get(step.id)) != null ? _b : []]);
+      const typoPool = [...strong, ...weak].filter((w) => !isSpacelessScript(w));
       let score = 0;
       const reasons = [];
-      for (const term of terms) {
+      for (const term of wordTerms) {
         if (strong.includes(term)) {
           score += 3;
           reasons.push({ kind: "keyword", term });
         } else if (weak.includes(term)) {
           score += 2;
           reasons.push({ kind: "title", term });
-        } else if ([...strong, ...weak].some((w) => withinOneEdit(term, w))) {
+        } else if (typoPool.some((w) => withinOneEdit(term, w))) {
           score += 0.5;
           reasons.push({ kind: "typo", term });
         }
       }
+      for (const hit of substringHits(strong, weak, haystack)) {
+        score += hit.weight;
+        reasons.push({ kind: hit.weight === 3 ? "keyword" : "title", term: hit.term });
+      }
       if (score > 0) candidates.push({ scriptId: script.id, stepId: step.id, score, reasons });
     }
   }
@@ -229,6 +312,18 @@ function matchQuestion(scripts, text, locale = "en") {
   if (!second || top.score >= second.score * 2) return { kind: "single", matches: [top] };
   return { kind: "multi", matches: candidates.filter((c) => c.score >= top.score / 2).slice(0, 3) };
 }
+function substringHits(strong, weak, haystack) {
+  const found = [];
+  const seen = /* @__PURE__ */ new Set();
+  for (const [list, weight] of [[strong, 3], [weak, 2]]) {
+    for (const term of list) {
+      if (seen.has(term) || !isSpacelessScript(term)) continue;
+      seen.add(term);
+      if (haystack.includes(term)) found.push({ term, weight });
+    }
+  }
+  return found.filter((a) => !found.some((b) => b.term !== a.term && b.weight >= a.weight && b.term.includes(a.term)));
+}
 
 // i18n/en.json
 var en_default = {
@@ -640,8 +735,49 @@ var tr_default = {
   "form.backToChat": "Sohbete d\xF6n"
 };
 
+// i18n/th.json
+var th_default = {
+  "chat.askAnswered": "\u0E04\u0E33\u0E15\u0E2D\u0E1A\u0E19\u0E35\u0E49\u0E0A\u0E48\u0E27\u0E22\u0E15\u0E2D\u0E1A\u0E04\u0E33\u0E16\u0E32\u0E21\u0E02\u0E2D\u0E07\u0E04\u0E38\u0E13\u0E2B\u0E23\u0E37\u0E2D\u0E44\u0E21\u0E48",
+  "chat.yes": "\u0E43\u0E0A\u0E48 \u0E02\u0E2D\u0E1A\u0E04\u0E38\u0E13",
+  "chat.no": "\u0E44\u0E21\u0E48\u0E43\u0E0A\u0E48 \u0E15\u0E49\u0E2D\u0E07\u0E01\u0E32\u0E23\u0E04\u0E27\u0E32\u0E21\u0E0A\u0E48\u0E27\u0E22\u0E40\u0E2B\u0E25\u0E37\u0E2D",
+  "chat.thanks": "\u0E22\u0E34\u0E19\u0E14\u0E35\u0E17\u0E35\u0E48\u0E44\u0E14\u0E49\u0E0A\u0E48\u0E27\u0E22\u0E40\u0E2B\u0E25\u0E37\u0E2D",
+  "chat.newQuestion": "\u0E16\u0E32\u0E21\u0E04\u0E33\u0E16\u0E32\u0E21\u0E43\u0E2B\u0E21\u0E48",
+  "chat.back": "\u0E22\u0E49\u0E2D\u0E19\u0E01\u0E25\u0E31\u0E1A",
+  "chat.home": "\u0E40\u0E21\u0E19\u0E39\u0E2B\u0E25\u0E31\u0E01",
+  "chat.didYouMean": "\u0E04\u0E38\u0E13\u0E2B\u0E21\u0E32\u0E22\u0E16\u0E36\u0E07\u0E02\u0E49\u0E2D\u0E43\u0E14",
+  "chat.notFound": "\u0E02\u0E2D\u0E2D\u0E20\u0E31\u0E22 \u0E44\u0E21\u0E48\u0E1E\u0E1A\u0E04\u0E33\u0E15\u0E2D\u0E1A\u0E2A\u0E33\u0E2B\u0E23\u0E31\u0E1A\u0E04\u0E33\u0E16\u0E32\u0E21\u0E19\u0E35\u0E49 \u0E01\u0E23\u0E38\u0E13\u0E32\u0E40\u0E25\u0E37\u0E2D\u0E01\u0E2B\u0E31\u0E27\u0E02\u0E49\u0E2D\u0E08\u0E32\u0E01\u0E40\u0E21\u0E19\u0E39 \u0E2B\u0E23\u0E37\u0E2D\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D\u0E17\u0E35\u0E21\u0E07\u0E32\u0E19\u0E02\u0E2D\u0E07\u0E40\u0E23\u0E32",
+  "chat.notFoundTwice": "\u0E02\u0E2D\u0E2D\u0E20\u0E31\u0E22 \u0E44\u0E21\u0E48\u0E1E\u0E1A\u0E04\u0E33\u0E15\u0E2D\u0E1A \u0E17\u0E35\u0E21\u0E07\u0E32\u0E19\u0E02\u0E2D\u0E07\u0E40\u0E23\u0E32\u0E22\u0E34\u0E19\u0E14\u0E35\u0E0A\u0E48\u0E27\u0E22\u0E40\u0E2B\u0E25\u0E37\u0E2D \u0E41\u0E25\u0E30\u0E44\u0E14\u0E49\u0E43\u0E2A\u0E48\u0E04\u0E33\u0E16\u0E32\u0E21\u0E02\u0E2D\u0E07\u0E04\u0E38\u0E13\u0E44\u0E27\u0E49\u0E43\u0E19\u0E41\u0E1A\u0E1A\u0E1F\u0E2D\u0E23\u0E4C\u0E21\u0E41\u0E25\u0E49\u0E27",
+  "chat.contactChip": "\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D\u0E17\u0E35\u0E21\u0E07\u0E32\u0E19",
+  "chat.openingContact": "\u0E01\u0E33\u0E25\u0E31\u0E07\u0E40\u0E1B\u0E34\u0E14\u0E41\u0E1A\u0E1A\u0E1F\u0E2D\u0E23\u0E4C\u0E21\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D \u0E1A\u0E17\u0E2A\u0E19\u0E17\u0E19\u0E32\u0E19\u0E35\u0E49\u0E08\u0E30\u0E41\u0E19\u0E1A\u0E44\u0E1B\u0E01\u0E31\u0E1A\u0E04\u0E33\u0E02\u0E2D\u0E02\u0E2D\u0E07\u0E04\u0E38\u0E13",
+  "chat.chooseTopic": "\u0E21\u0E35\u0E2D\u0E30\u0E44\u0E23\u0E43\u0E2B\u0E49\u0E0A\u0E48\u0E27\u0E22\u0E44\u0E2B\u0E21",
+  "chat.aiLabel": "\u0E04\u0E33\u0E15\u0E2D\u0E1A\u0E08\u0E32\u0E01 AI",
+  "widget.open": "\u0E40\u0E1B\u0E34\u0E14\u0E41\u0E0A\u0E15",
+  "widget.close": "\u0E1B\u0E34\u0E14\u0E41\u0E0A\u0E15",
+  "widget.placeholder": "\u0E1E\u0E34\u0E21\u0E1E\u0E4C\u0E04\u0E33\u0E16\u0E32\u0E21\u0E02\u0E2D\u0E07\u0E04\u0E38\u0E13...",
+  "widget.send": "\u0E2A\u0E48\u0E07",
+  "widget.poweredBy": "\u0E02\u0E31\u0E1A\u0E40\u0E04\u0E25\u0E37\u0E48\u0E2D\u0E19\u0E42\u0E14\u0E22 Kavchat",
+  "widget.offline": "\u0E04\u0E38\u0E13\u0E2D\u0E2D\u0E1F\u0E44\u0E25\u0E19\u0E4C\u0E2D\u0E22\u0E39\u0E48 \u0E01\u0E33\u0E25\u0E31\u0E07\u0E41\u0E2A\u0E14\u0E07\u0E04\u0E33\u0E15\u0E2D\u0E1A\u0E17\u0E35\u0E48\u0E1A\u0E31\u0E19\u0E17\u0E36\u0E01\u0E44\u0E27\u0E49",
+  "widget.loading": "\u0E01\u0E33\u0E25\u0E31\u0E07\u0E42\u0E2B\u0E25\u0E14...",
+  "widget.unavailable": "\u0E1C\u0E39\u0E49\u0E0A\u0E48\u0E27\u0E22\u0E44\u0E21\u0E48\u0E1E\u0E23\u0E49\u0E2D\u0E21\u0E43\u0E2B\u0E49\u0E1A\u0E23\u0E34\u0E01\u0E32\u0E23\u0E43\u0E19\u0E02\u0E13\u0E30\u0E19\u0E35\u0E49",
+  "form.title": "\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D\u0E40\u0E23\u0E32",
+  "form.name": "\u0E0A\u0E37\u0E48\u0E2D",
+  "form.email": "\u0E2D\u0E35\u0E40\u0E21\u0E25",
+  "form.phone": "\u0E42\u0E17\u0E23\u0E28\u0E31\u0E1E\u0E17\u0E4C",
+  "form.topic": "\u0E2B\u0E31\u0E27\u0E02\u0E49\u0E2D",
+  "form.message": "\u0E02\u0E49\u0E2D\u0E04\u0E27\u0E32\u0E21",
+  "form.submit": "\u0E2A\u0E48\u0E07",
+  "form.cancel": "\u0E22\u0E01\u0E40\u0E25\u0E34\u0E01",
+  "form.sending": "\u0E01\u0E33\u0E25\u0E31\u0E07\u0E2A\u0E48\u0E07...",
+  "form.sent": "\u0E02\u0E2D\u0E1A\u0E04\u0E38\u0E13 \u0E40\u0E23\u0E32\u0E44\u0E14\u0E49\u0E23\u0E31\u0E1A\u0E04\u0E33\u0E02\u0E2D\u0E2B\u0E21\u0E32\u0E22\u0E40\u0E25\u0E02 {number} \u0E02\u0E2D\u0E07\u0E04\u0E38\u0E13\u0E41\u0E25\u0E49\u0E27 \u0E41\u0E25\u0E30\u0E08\u0E30\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D\u0E01\u0E25\u0E31\u0E1A\u0E42\u0E14\u0E22\u0E40\u0E23\u0E47\u0E27",
+  "form.error": "\u0E40\u0E01\u0E34\u0E14\u0E02\u0E49\u0E2D\u0E1C\u0E34\u0E14\u0E1E\u0E25\u0E32\u0E14 \u0E01\u0E23\u0E38\u0E13\u0E32\u0E25\u0E2D\u0E07\u0E2D\u0E35\u0E01\u0E04\u0E23\u0E31\u0E49\u0E07",
+  "form.required": "\u0E08\u0E33\u0E40\u0E1B\u0E47\u0E19\u0E15\u0E49\u0E2D\u0E07\u0E01\u0E23\u0E2D\u0E01",
+  "form.contactRequired": "\u0E01\u0E23\u0E38\u0E13\u0E32\u0E23\u0E30\u0E1A\u0E38\u0E2D\u0E35\u0E40\u0E21\u0E25\u0E2B\u0E23\u0E37\u0E2D\u0E2B\u0E21\u0E32\u0E22\u0E40\u0E25\u0E02\u0E42\u0E17\u0E23\u0E28\u0E31\u0E1E\u0E17\u0E4C \u0E40\u0E1E\u0E37\u0E48\u0E2D\u0E43\u0E2B\u0E49\u0E40\u0E23\u0E32\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D\u0E01\u0E25\u0E31\u0E1A\u0E44\u0E14\u0E49",
+  "form.emailInvalid": "\u0E2D\u0E35\u0E40\u0E21\u0E25\u0E44\u0E21\u0E48\u0E16\u0E39\u0E01\u0E15\u0E49\u0E2D\u0E07",
+  "form.backToChat": "\u0E01\u0E25\u0E31\u0E1A\u0E44\u0E1B\u0E17\u0E35\u0E48\u0E41\u0E0A\u0E15"
+};
+
 // src/i18n.ts
-var UI_TEXTS = { en: en_default, he: he_default, ar: ar_default, ru: ru_default, es: es_default, fr: fr_default, de: de_default, pt: pt_default, it: it_default, tr: tr_default };
+var UI_TEXTS = { en: en_default, he: he_default, ar: ar_default, ru: ru_default, es: es_default, fr: fr_default, de: de_default, pt: pt_default, it: it_default, tr: tr_default, th: th_default };
 var RTL_LOCALES = ["he", "ar"];
 var isRtl = (locale) => RTL_LOCALES.includes(baseLocale(locale));
 var baseLocale = (locale) => String(locale || "en").toLowerCase().split(/[-_]/)[0].replace(/^iw$/, "he");
@@ -925,6 +1061,7 @@ function eventsPayload(state, sessionId, releaseVersion, startedAt) {
   format,
   groupKey,
   isRtl,
+  isSpacelessScript,
   keyTerms,
   matchQuestion,
   normalizeWords,

package/dist/index.mjs

@@ -1,5 +1,5 @@
 // src/normalize.ts
-var SUPPORTED_LOCALES = ["he", "en", "ar", "ru", "es", "fr", "de", "pt", "it", "tr"];
+var SUPPORTED_LOCALES = ["he", "en", "ar", "ru", "es", "fr", "de", "pt", "it", "tr", "th"];
 var FOLD = {
   // עברית: אותיות סופיות
   \u05DA: "\u05DB",
@@ -25,6 +25,13 @@ var FOLD = {
 function isCombiningMark(code) {
   return code >= 768 && code <= 879 || code >= 1425 && code <= 1479 || code >= 1552 && code <= 1562 || code >= 1611 && code <= 1631 || code === 1648 || code >= 1750 && code <= 1773;
 }
+function isSpacelessLetter(code) {
+  return code >= 3585 && code <= 3630 || code >= 3632 && code <= 3642 || code >= 3648 && code <= 3653 || code >= 3655 && code <= 3662 || code >= 3713 && code <= 3758 || code >= 3760 && code <= 3781 || code >= 3784 && code <= 3790 || code >= 3804 && code <= 3807 || code >= 4096 && code <= 4169 || code >= 4176 && code <= 4253 || code >= 6016 && code <= 6099 || code === 6108 || code === 6109 || code === 12293 || code >= 12353 && code <= 12438 || code >= 12441 && code <= 12442 || code >= 12445 && code <= 12447 || code >= 12449 && code <= 12538 || code >= 12540 && code <= 12543 || code >= 13312 && code <= 19903 || code >= 19968 && code <= 40959;
+}
+function isSpacelessScript(word) {
+  for (let i = 0; i < word.length; i++) if (isSpacelessLetter(word.charCodeAt(i))) return true;
+  return false;
+}
 function isWordChar(ch) {
   const code = ch.charCodeAt(0);
   if (code >= 48 && code <= 57) return true;
@@ -32,9 +39,75 @@ function isWordChar(ch) {
   if (code >= 1488 && code <= 1514) return true;
   if (code >= 1568 && code <= 1610) return true;
   if (code >= 1649 && code <= 1747) return true;
-  if (code >= 19968 && code <= 40959) return true;
+  if (isSpacelessLetter(code)) return true;
   return ch.toLowerCase() !== ch.toUpperCase();
 }
+var isZeroWidth = (code) => code >= 8203 && code <= 8205 || code === 8288 || code === 65279;
+var DIGIT_ZEROS = [3664, 3792, 4160, 6112];
+function asciiDigit(code) {
+  for (const zero of DIGIT_ZEROS) if (code >= zero && code <= zero + 9) return String.fromCharCode(48 + code - zero);
+  return void 0;
+}
+var MARK_ORDER = {
+  3640: 103,
+  3641: 103,
+  3642: 9,
+  3656: 107,
+  3657: 107,
+  3658: 107,
+  3659: 107,
+  3768: 118,
+  3769: 118,
+  3770: 9,
+  3784: 122,
+  3785: 122,
+  3786: 122,
+  3787: 122,
+  4151: 7,
+  4153: 9,
+  4154: 9,
+  4237: 220,
+  6098: 9,
+  6109: 230,
+  12441: 8,
+  12442: 8
+};
+var isThaiTone = (code) => code >= 3656 && code <= 3659;
+function orderMarks(text) {
+  if (!/[\u0e33\u0e38-\u0e3a\u0e48-\u0e4b\u0e4d\u0eb8-\u0eba\u0ec8-\u0ecb\u1037\u1039\u103a\u108d\u17d2\u17dd\u3099\u309a]/.test(text)) return text;
+  const units = [];
+  for (let i = 0; i < text.length; i++) units.push(text.charCodeAt(i));
+  for (let i = 0; i < units.length; i++) {
+    if (!MARK_ORDER[units[i]]) continue;
+    let end = i;
+    while (end < units.length && MARK_ORDER[units[end]]) end++;
+    for (let a = i + 1; a < end; a++) {
+      const v = units[a];
+      let b = a - 1;
+      while (b >= i && MARK_ORDER[units[b]] > MARK_ORDER[v]) {
+        units[b + 1] = units[b];
+        b--;
+      }
+      units[b + 1] = v;
+    }
+    i = end - 1;
+  }
+  let out = "";
+  for (let i = 0; i < units.length; i++) {
+    const c = units[i];
+    if (c === 3661 && units[i + 1] === 3634) {
+      out += "\u0E33";
+      i++;
+    } else if (c === 3661 && isThaiTone(units[i + 1]) && units[i + 2] === 3634) {
+      out += String.fromCharCode(units[i + 1], 3635);
+      i += 2;
+    } else if (c === 3635 && isThaiTone(units[i + 1])) {
+      out += String.fromCharCode(units[i + 1], 3635);
+      i++;
+    } else out += String.fromCharCode(c);
+  }
+  return out;
+}
 var decompose = (s) => {
   try {
     return s.normalize("NFD");
@@ -43,16 +116,16 @@ var decompose = (s) => {
   }
 };
 function normalizeWords(text) {
-  var _a;
+  var _a, _b;
   const source = decompose(String(text != null ? text : "").toLowerCase());
   let out = "";
   for (const raw of source) {
     const code = raw.charCodeAt(0);
-    if (isCombiningMark(code)) continue;
-    const ch = (_a = FOLD[raw]) != null ? _a : raw;
+    if (isCombiningMark(code) || isZeroWidth(code)) continue;
+    const ch = (_b = (_a = FOLD[raw]) != null ? _a : asciiDigit(code)) != null ? _b : raw;
     out += ch.length === 1 && !isWordChar(ch) ? " " : ch;
   }
-  const cleaned = out.replace(/\s+/g, " ").trim();
+  const cleaned = orderMarks(out).replace(/\s+/g, " ").trim();
   return cleaned ? cleaned.split(" ") : [];
 }
 var STOP_WORDS = {
@@ -65,7 +138,9 @@ var STOP_WORDS = {
   de: ["wie", "was", "ich", "mein", "meine", "mir", "der", "die", "das", "ein", "eine", "und", "oder", "in", "mit", "fur", "von", "zu", "ist", "kann", "mochte", "wann", "wo", "warum", "nicht", "es", "den", "dem"],
   pt: ["como", "que", "eu", "meu", "minha", "o", "a", "os", "as", "um", "uma", "de", "do", "da", "em", "no", "na", "com", "para", "por", "e", "ou", "posso", "quero", "quando", "onde", "porque", "nao"],
   it: ["come", "che", "cosa", "io", "mio", "mia", "il", "lo", "la", "i", "gli", "le", "un", "una", "di", "del", "della", "in", "con", "per", "e", "o", "posso", "voglio", "quando", "dove", "perche", "non", "l"],
-  tr: ["nasil", "ne", "ben", "benim", "bir", "ve", "veya", "ile", "icin", "bu", "su", "mi", "mu", "nerede", "ne zaman", "neden", "kac", "istiyorum", "var", "yok", "de", "da"]
+  tr: ["nasil", "ne", "ben", "benim", "bir", "ve", "veya", "ile", "icin", "bu", "su", "mi", "mu", "nerede", "ne zaman", "neden", "kac", "istiyorum", "var", "yok", "de", "da"],
+  // בתאית מילות קישור נמחקות רק כשהן מילה נפרדת (מוקפות ברווחים). רוב השאלות בלי רווחים, וההתאמה שם לפי תת-מחרוזת
+  th: ["\u0E09\u0E31\u0E19", "\u0E1C\u0E21", "\u0E14\u0E34\u0E09\u0E31\u0E19", "\u0E04\u0E38\u0E13", "\u0E04\u0E23\u0E31\u0E1A", "\u0E04\u0E48\u0E30", "\u0E04\u0E30", "\u0E19\u0E30", "\u0E44\u0E2B\u0E21", "\u0E21\u0E31\u0E49\u0E22", "\u0E2D\u0E30\u0E44\u0E23", "\u0E2D\u0E22\u0E48\u0E32\u0E07\u0E44\u0E23", "\u0E22\u0E31\u0E07\u0E44\u0E07", "\u0E17\u0E35\u0E48", "\u0E41\u0E25\u0E30", "\u0E2B\u0E23\u0E37\u0E2D", "\u0E02\u0E2D\u0E07", "\u0E44\u0E14\u0E49", "\u0E08\u0E30", "\u0E43\u0E19", "\u0E01\u0E31\u0E1A", "\u0E40\u0E1B\u0E47\u0E19", "\u0E21\u0E35", "\u0E44\u0E21\u0E48", "\u0E43\u0E2B\u0E49", "\u0E27\u0E48\u0E32", "\u0E2D\u0E22\u0E32\u0E01", "\u0E15\u0E49\u0E2D\u0E07\u0E01\u0E32\u0E23", "\u0E40\u0E21\u0E37\u0E48\u0E2D\u0E44\u0E2B\u0E23\u0E48", "\u0E17\u0E35\u0E48\u0E44\u0E2B\u0E19", "\u0E17\u0E33\u0E44\u0E21", "\u0E40\u0E17\u0E48\u0E32\u0E44\u0E2B\u0E23\u0E48"]
 };
 var NORMALIZED_STOP_WORDS = {};
 for (const locale of SUPPORTED_LOCALES) {
@@ -140,6 +215,8 @@ function matchQuestion(scripts, text, locale = "en") {
   const terms = Array.from(new Set(keyTerms(text, locale)));
   if (terms.length === 0) return { kind: "none", matches: [] };
   const terms_ = (values) => values.flatMap((v) => keyTerms(v, locale));
+  const wordTerms = terms.filter((t) => !isSpacelessScript(t));
+  const haystack = normalizeWords(text).join("");
   const candidates = [];
   for (const script of scripts) {
     if (script.kind === "opening") continue;
@@ -157,20 +234,25 @@ function matchQuestion(scripts, text, locale = "en") {
       const isRoot = step.id === content.rootStepId;
       const strong = terms_([...step.keywords, ...isRoot ? content.keywords : []]);
       const weak = terms_([...isRoot ? [content.name] : [], ...(_b = labelsTo.get(step.id)) != null ? _b : []]);
+      const typoPool = [...strong, ...weak].filter((w) => !isSpacelessScript(w));
       let score = 0;
       const reasons = [];
-      for (const term of terms) {
+      for (const term of wordTerms) {
         if (strong.includes(term)) {
           score += 3;
           reasons.push({ kind: "keyword", term });
         } else if (weak.includes(term)) {
           score += 2;
           reasons.push({ kind: "title", term });
-        } else if ([...strong, ...weak].some((w) => withinOneEdit(term, w))) {
+        } else if (typoPool.some((w) => withinOneEdit(term, w))) {
           score += 0.5;
           reasons.push({ kind: "typo", term });
         }
       }
+      for (const hit of substringHits(strong, weak, haystack)) {
+        score += hit.weight;
+        reasons.push({ kind: hit.weight === 3 ? "keyword" : "title", term: hit.term });
+      }
       if (score > 0) candidates.push({ scriptId: script.id, stepId: step.id, score, reasons });
     }
   }
@@ -180,6 +262,18 @@ function matchQuestion(scripts, text, locale = "en") {
   if (!second || top.score >= second.score * 2) return { kind: "single", matches: [top] };
   return { kind: "multi", matches: candidates.filter((c) => c.score >= top.score / 2).slice(0, 3) };
 }
+function substringHits(strong, weak, haystack) {
+  const found = [];
+  const seen = /* @__PURE__ */ new Set();
+  for (const [list, weight] of [[strong, 3], [weak, 2]]) {
+    for (const term of list) {
+      if (seen.has(term) || !isSpacelessScript(term)) continue;
+      seen.add(term);
+      if (haystack.includes(term)) found.push({ term, weight });
+    }
+  }
+  return found.filter((a) => !found.some((b) => b.term !== a.term && b.weight >= a.weight && b.term.includes(a.term)));
+}
 
 // i18n/en.json
 var en_default = {
@@ -591,8 +685,49 @@ var tr_default = {
   "form.backToChat": "Sohbete d\xF6n"
 };
 
+// i18n/th.json
+var th_default = {
+  "chat.askAnswered": "\u0E04\u0E33\u0E15\u0E2D\u0E1A\u0E19\u0E35\u0E49\u0E0A\u0E48\u0E27\u0E22\u0E15\u0E2D\u0E1A\u0E04\u0E33\u0E16\u0E32\u0E21\u0E02\u0E2D\u0E07\u0E04\u0E38\u0E13\u0E2B\u0E23\u0E37\u0E2D\u0E44\u0E21\u0E48",
+  "chat.yes": "\u0E43\u0E0A\u0E48 \u0E02\u0E2D\u0E1A\u0E04\u0E38\u0E13",
+  "chat.no": "\u0E44\u0E21\u0E48\u0E43\u0E0A\u0E48 \u0E15\u0E49\u0E2D\u0E07\u0E01\u0E32\u0E23\u0E04\u0E27\u0E32\u0E21\u0E0A\u0E48\u0E27\u0E22\u0E40\u0E2B\u0E25\u0E37\u0E2D",
+  "chat.thanks": "\u0E22\u0E34\u0E19\u0E14\u0E35\u0E17\u0E35\u0E48\u0E44\u0E14\u0E49\u0E0A\u0E48\u0E27\u0E22\u0E40\u0E2B\u0E25\u0E37\u0E2D",
+  "chat.newQuestion": "\u0E16\u0E32\u0E21\u0E04\u0E33\u0E16\u0E32\u0E21\u0E43\u0E2B\u0E21\u0E48",
+  "chat.back": "\u0E22\u0E49\u0E2D\u0E19\u0E01\u0E25\u0E31\u0E1A",
+  "chat.home": "\u0E40\u0E21\u0E19\u0E39\u0E2B\u0E25\u0E31\u0E01",
+  "chat.didYouMean": "\u0E04\u0E38\u0E13\u0E2B\u0E21\u0E32\u0E22\u0E16\u0E36\u0E07\u0E02\u0E49\u0E2D\u0E43\u0E14",
+  "chat.notFound": "\u0E02\u0E2D\u0E2D\u0E20\u0E31\u0E22 \u0E44\u0E21\u0E48\u0E1E\u0E1A\u0E04\u0E33\u0E15\u0E2D\u0E1A\u0E2A\u0E33\u0E2B\u0E23\u0E31\u0E1A\u0E04\u0E33\u0E16\u0E32\u0E21\u0E19\u0E35\u0E49 \u0E01\u0E23\u0E38\u0E13\u0E32\u0E40\u0E25\u0E37\u0E2D\u0E01\u0E2B\u0E31\u0E27\u0E02\u0E49\u0E2D\u0E08\u0E32\u0E01\u0E40\u0E21\u0E19\u0E39 \u0E2B\u0E23\u0E37\u0E2D\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D\u0E17\u0E35\u0E21\u0E07\u0E32\u0E19\u0E02\u0E2D\u0E07\u0E40\u0E23\u0E32",
+  "chat.notFoundTwice": "\u0E02\u0E2D\u0E2D\u0E20\u0E31\u0E22 \u0E44\u0E21\u0E48\u0E1E\u0E1A\u0E04\u0E33\u0E15\u0E2D\u0E1A \u0E17\u0E35\u0E21\u0E07\u0E32\u0E19\u0E02\u0E2D\u0E07\u0E40\u0E23\u0E32\u0E22\u0E34\u0E19\u0E14\u0E35\u0E0A\u0E48\u0E27\u0E22\u0E40\u0E2B\u0E25\u0E37\u0E2D \u0E41\u0E25\u0E30\u0E44\u0E14\u0E49\u0E43\u0E2A\u0E48\u0E04\u0E33\u0E16\u0E32\u0E21\u0E02\u0E2D\u0E07\u0E04\u0E38\u0E13\u0E44\u0E27\u0E49\u0E43\u0E19\u0E41\u0E1A\u0E1A\u0E1F\u0E2D\u0E23\u0E4C\u0E21\u0E41\u0E25\u0E49\u0E27",
+  "chat.contactChip": "\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D\u0E17\u0E35\u0E21\u0E07\u0E32\u0E19",
+  "chat.openingContact": "\u0E01\u0E33\u0E25\u0E31\u0E07\u0E40\u0E1B\u0E34\u0E14\u0E41\u0E1A\u0E1A\u0E1F\u0E2D\u0E23\u0E4C\u0E21\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D \u0E1A\u0E17\u0E2A\u0E19\u0E17\u0E19\u0E32\u0E19\u0E35\u0E49\u0E08\u0E30\u0E41\u0E19\u0E1A\u0E44\u0E1B\u0E01\u0E31\u0E1A\u0E04\u0E33\u0E02\u0E2D\u0E02\u0E2D\u0E07\u0E04\u0E38\u0E13",
+  "chat.chooseTopic": "\u0E21\u0E35\u0E2D\u0E30\u0E44\u0E23\u0E43\u0E2B\u0E49\u0E0A\u0E48\u0E27\u0E22\u0E44\u0E2B\u0E21",
+  "chat.aiLabel": "\u0E04\u0E33\u0E15\u0E2D\u0E1A\u0E08\u0E32\u0E01 AI",
+  "widget.open": "\u0E40\u0E1B\u0E34\u0E14\u0E41\u0E0A\u0E15",
+  "widget.close": "\u0E1B\u0E34\u0E14\u0E41\u0E0A\u0E15",
+  "widget.placeholder": "\u0E1E\u0E34\u0E21\u0E1E\u0E4C\u0E04\u0E33\u0E16\u0E32\u0E21\u0E02\u0E2D\u0E07\u0E04\u0E38\u0E13...",
+  "widget.send": "\u0E2A\u0E48\u0E07",
+  "widget.poweredBy": "\u0E02\u0E31\u0E1A\u0E40\u0E04\u0E25\u0E37\u0E48\u0E2D\u0E19\u0E42\u0E14\u0E22 Kavchat",
+  "widget.offline": "\u0E04\u0E38\u0E13\u0E2D\u0E2D\u0E1F\u0E44\u0E25\u0E19\u0E4C\u0E2D\u0E22\u0E39\u0E48 \u0E01\u0E33\u0E25\u0E31\u0E07\u0E41\u0E2A\u0E14\u0E07\u0E04\u0E33\u0E15\u0E2D\u0E1A\u0E17\u0E35\u0E48\u0E1A\u0E31\u0E19\u0E17\u0E36\u0E01\u0E44\u0E27\u0E49",
+  "widget.loading": "\u0E01\u0E33\u0E25\u0E31\u0E07\u0E42\u0E2B\u0E25\u0E14...",
+  "widget.unavailable": "\u0E1C\u0E39\u0E49\u0E0A\u0E48\u0E27\u0E22\u0E44\u0E21\u0E48\u0E1E\u0E23\u0E49\u0E2D\u0E21\u0E43\u0E2B\u0E49\u0E1A\u0E23\u0E34\u0E01\u0E32\u0E23\u0E43\u0E19\u0E02\u0E13\u0E30\u0E19\u0E35\u0E49",
+  "form.title": "\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D\u0E40\u0E23\u0E32",
+  "form.name": "\u0E0A\u0E37\u0E48\u0E2D",
+  "form.email": "\u0E2D\u0E35\u0E40\u0E21\u0E25",
+  "form.phone": "\u0E42\u0E17\u0E23\u0E28\u0E31\u0E1E\u0E17\u0E4C",
+  "form.topic": "\u0E2B\u0E31\u0E27\u0E02\u0E49\u0E2D",
+  "form.message": "\u0E02\u0E49\u0E2D\u0E04\u0E27\u0E32\u0E21",
+  "form.submit": "\u0E2A\u0E48\u0E07",
+  "form.cancel": "\u0E22\u0E01\u0E40\u0E25\u0E34\u0E01",
+  "form.sending": "\u0E01\u0E33\u0E25\u0E31\u0E07\u0E2A\u0E48\u0E07...",
+  "form.sent": "\u0E02\u0E2D\u0E1A\u0E04\u0E38\u0E13 \u0E40\u0E23\u0E32\u0E44\u0E14\u0E49\u0E23\u0E31\u0E1A\u0E04\u0E33\u0E02\u0E2D\u0E2B\u0E21\u0E32\u0E22\u0E40\u0E25\u0E02 {number} \u0E02\u0E2D\u0E07\u0E04\u0E38\u0E13\u0E41\u0E25\u0E49\u0E27 \u0E41\u0E25\u0E30\u0E08\u0E30\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D\u0E01\u0E25\u0E31\u0E1A\u0E42\u0E14\u0E22\u0E40\u0E23\u0E47\u0E27",
+  "form.error": "\u0E40\u0E01\u0E34\u0E14\u0E02\u0E49\u0E2D\u0E1C\u0E34\u0E14\u0E1E\u0E25\u0E32\u0E14 \u0E01\u0E23\u0E38\u0E13\u0E32\u0E25\u0E2D\u0E07\u0E2D\u0E35\u0E01\u0E04\u0E23\u0E31\u0E49\u0E07",
+  "form.required": "\u0E08\u0E33\u0E40\u0E1B\u0E47\u0E19\u0E15\u0E49\u0E2D\u0E07\u0E01\u0E23\u0E2D\u0E01",
+  "form.contactRequired": "\u0E01\u0E23\u0E38\u0E13\u0E32\u0E23\u0E30\u0E1A\u0E38\u0E2D\u0E35\u0E40\u0E21\u0E25\u0E2B\u0E23\u0E37\u0E2D\u0E2B\u0E21\u0E32\u0E22\u0E40\u0E25\u0E02\u0E42\u0E17\u0E23\u0E28\u0E31\u0E1E\u0E17\u0E4C \u0E40\u0E1E\u0E37\u0E48\u0E2D\u0E43\u0E2B\u0E49\u0E40\u0E23\u0E32\u0E15\u0E34\u0E14\u0E15\u0E48\u0E2D\u0E01\u0E25\u0E31\u0E1A\u0E44\u0E14\u0E49",
+  "form.emailInvalid": "\u0E2D\u0E35\u0E40\u0E21\u0E25\u0E44\u0E21\u0E48\u0E16\u0E39\u0E01\u0E15\u0E49\u0E2D\u0E07",
+  "form.backToChat": "\u0E01\u0E25\u0E31\u0E1A\u0E44\u0E1B\u0E17\u0E35\u0E48\u0E41\u0E0A\u0E15"
+};
+
 // src/i18n.ts
-var UI_TEXTS = { en: en_default, he: he_default, ar: ar_default, ru: ru_default, es: es_default, fr: fr_default, de: de_default, pt: pt_default, it: it_default, tr: tr_default };
+var UI_TEXTS = { en: en_default, he: he_default, ar: ar_default, ru: ru_default, es: es_default, fr: fr_default, de: de_default, pt: pt_default, it: it_default, tr: tr_default, th: th_default };
 var RTL_LOCALES = ["he", "ar"];
 var isRtl = (locale) => RTL_LOCALES.includes(baseLocale(locale));
 var baseLocale = (locale) => String(locale || "en").toLowerCase().split(/[-_]/)[0].replace(/^iw$/, "he");
@@ -875,6 +1010,7 @@ export {
   format,
   groupKey,
   isRtl,
+  isSpacelessScript,
   keyTerms,
   matchQuestion,
   normalizeWords,

package/i18n/th.json

@@ -0,0 +1,39 @@
+{
+  "chat.askAnswered": "คำตอบนี้ช่วยตอบคำถามของคุณหรือไม่",
+  "chat.yes": "ใช่ ขอบคุณ",
+  "chat.no": "ไม่ใช่ ต้องการความช่วยเหลือ",
+  "chat.thanks": "ยินดีที่ได้ช่วยเหลือ",
+  "chat.newQuestion": "ถามคำถามใหม่",
+  "chat.back": "ย้อนกลับ",
+  "chat.home": "เมนูหลัก",
+  "chat.didYouMean": "คุณหมายถึงข้อใด",
+  "chat.notFound": "ขออภัย ไม่พบคำตอบสำหรับคำถามนี้ กรุณาเลือกหัวข้อจากเมนู หรือติดต่อทีมงานของเรา",
+  "chat.notFoundTwice": "ขออภัย ไม่พบคำตอบ ทีมงานของเรายินดีช่วยเหลือ และได้ใส่คำถามของคุณไว้ในแบบฟอร์มแล้ว",
+  "chat.contactChip": "ติดต่อทีมงาน",
+  "chat.openingContact": "กำลังเปิดแบบฟอร์มติดต่อ บทสนทนานี้จะแนบไปกับคำขอของคุณ",
+  "chat.chooseTopic": "มีอะไรให้ช่วยไหม",
+  "chat.aiLabel": "คำตอบจาก AI",
+  "widget.open": "เปิดแชต",
+  "widget.close": "ปิดแชต",
+  "widget.placeholder": "พิมพ์คำถามของคุณ...",
+  "widget.send": "ส่ง",
+  "widget.poweredBy": "ขับเคลื่อนโดย Kavchat",
+  "widget.offline": "คุณออฟไลน์อยู่ กำลังแสดงคำตอบที่บันทึกไว้",
+  "widget.loading": "กำลังโหลด...",
+  "widget.unavailable": "ผู้ช่วยไม่พร้อมให้บริการในขณะนี้",
+  "form.title": "ติดต่อเรา",
+  "form.name": "ชื่อ",
+  "form.email": "อีเมล",
+  "form.phone": "โทรศัพท์",
+  "form.topic": "หัวข้อ",
+  "form.message": "ข้อความ",
+  "form.submit": "ส่ง",
+  "form.cancel": "ยกเลิก",
+  "form.sending": "กำลังส่ง...",
+  "form.sent": "ขอบคุณ เราได้รับคำขอหมายเลข {number} ของคุณแล้ว และจะติดต่อกลับโดยเร็ว",
+  "form.error": "เกิดข้อผิดพลาด กรุณาลองอีกครั้ง",
+  "form.required": "จำเป็นต้องกรอก",
+  "form.contactRequired": "กรุณาระบุอีเมลหรือหมายเลขโทรศัพท์ เพื่อให้เราติดต่อกลับได้",
+  "form.emailInvalid": "อีเมลไม่ถูกต้อง",
+  "form.backToChat": "กลับไปที่แชต"
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kavchat/core",
-  "version": "0.0.1",
+  "version": "0.0.2",
   "description": "Kavchat conversation engine: text normalization, question matching and the chat state machine. Zero dependencies.",
   "license": "MIT",
   "main": "dist/index.js",