ssml-builder-js 2.3.0 → 2.5.0

package/dist/{chunk-PSN6XAFF.mjs → chunk-4RHHW33A.mjs}

@@ -848,10 +848,559 @@ function validateSsml(xmlString) {
   }
 }
 
+// packages/ssml-core/src/textNodes.ts
+function decodeXmlText(value) {
+  return value.replace(/&(?:amp|apos|gt|lt|quot);|&#(?:x[\da-f]+|\d+);/gi, (entity) => {
+    if (entity === "&") return "&";
+    if (entity === "'") return "'";
+    if (entity === ">") return ">";
+    if (entity === "&lt;") return "<";
+    if (entity === "&quot;") return '"';
+    const hexadecimal = entity.toLowerCase().startsWith("&#x");
+    const digits = entity.slice(hexadecimal ? 3 : 2, -1);
+    return String.fromCodePoint(Number.parseInt(digits, hexadecimal ? 16 : 10));
+  });
+}
+function encodeXmlText(value) {
+  return value.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;");
+}
+function decodeXmlAttribute(value) {
+  return decodeXmlText(value);
+}
+function findTagEnd(source, start) {
+  let quote = "";
+  for (let index = start; index < source.length; index += 1) {
+    const character = source[index];
+    if (quote) {
+      if (character === quote) quote = "";
+    } else if (character === '"' || character === "'") {
+      quote = character;
+    } else if (character === ">") {
+      return index;
+    }
+  }
+  return source.length - 1;
+}
+function readTagName(tag) {
+  const match = /^<\s*([A-Za-z_][A-Za-z0-9_.:-]*)/.exec(tag);
+  return match?.[1];
+}
+function readTagAttributes(tag, name) {
+  const attributes = {};
+  const nameStart = tag.indexOf(name);
+  const attributeSource = tag.slice(nameStart + name.length, tag.length - 1).replace(/\/\s*$/, "");
+  const attributePattern = /([A-Za-z_][A-Za-z0-9_.:-]*)\s*=\s*(["'])([\s\S]*?)\2/g;
+  for (const match of attributeSource.matchAll(attributePattern)) {
+    attributes[match[1].toLowerCase()] = decodeXmlAttribute(match[3]);
+  }
+  return attributes;
+}
+function collectTextNodes(source) {
+  const nodes = [];
+  const elements = [];
+  let index = 0;
+  const addText = (start, end, rawText, sourceStart = start, sourceEnd = end) => {
+    if (!rawText) return;
+    const path = elements.map((element) => element.name);
+    const parent = elements[elements.length - 1];
+    nodes.push({
+      context: {
+        ancestorTags: path.slice(0, -1),
+        parentAttributes: { ...parent?.attributes ?? {} },
+        parentTag: parent?.name ?? "",
+        path
+      },
+      decodedText: decodeXmlText(rawText),
+      end,
+      sourceEnd,
+      sourceStart,
+      start
+    });
+  };
+  while (index < source.length) {
+    if (source[index] !== "<") {
+      const nextTag = source.indexOf("<", index);
+      const end2 = nextTag === -1 ? source.length : nextTag;
+      addText(index, end2, source.slice(index, end2));
+      index = end2;
+      continue;
+    }
+    if (source.startsWith("<!--", index)) {
+      const end2 = source.indexOf("-->", index + 4);
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<![CDATA[", index)) {
+      const contentStart = index + 9;
+      const end2 = source.indexOf("]]>", contentStart);
+      const contentEnd = end2 === -1 ? source.length : end2;
+      addText(
+        contentStart,
+        contentEnd,
+        source.slice(contentStart, contentEnd),
+        index,
+        end2 === -1 ? source.length : end2 + 3
+      );
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<?", index)) {
+      const end2 = source.indexOf("?>", index + 2);
+      index = end2 === -1 ? source.length : end2 + 2;
+      continue;
+    }
+    if (source.startsWith("</", index)) {
+      const end2 = findTagEnd(source, index + 2);
+      elements.pop();
+      index = end2 + 1;
+      continue;
+    }
+    const end = findTagEnd(source, index + 1);
+    const tag = source.slice(index, end + 1);
+    const name = readTagName(tag);
+    if (name && !/\/\s*>$/.test(tag)) elements.push({ attributes: readTagAttributes(tag, name), name });
+    index = end + 1;
+  }
+  return nodes;
+}
+function extractSsmlText(ssml) {
+  parseSsml(ssml);
+  return collectTextNodes(ssml).map((node) => node.decodedText);
+}
+async function mapSsmlTextNodes(ssml, transform, options = {}) {
+  parseSsml(ssml);
+  const nodes = collectTextNodes(ssml);
+  const skipTags = new Set((options.skipTags ?? ["phoneme", "say-as", "sub"]).map((tag) => tag.toLowerCase()));
+  const replacements = await Promise.all(
+    nodes.map(async (node) => {
+      const context = {
+        ancestorTags: [...node.context.ancestorTags],
+        parentAttributes: { ...node.context.parentAttributes },
+        parentTag: node.context.parentTag,
+        path: [...node.context.path]
+      };
+      const shouldTransform = !skipTags.has(context.parentTag.toLowerCase()) && (options.filter?.(context) ?? true);
+      if (!shouldTransform) return ssml.slice(node.sourceStart, node.sourceEnd);
+      const transformed = await transform(node.decodedText, context);
+      if (typeof transformed !== "string") {
+        throw new TypeError("SSML text node transform must return a string");
+      }
+      return transformed === node.decodedText ? ssml.slice(node.sourceStart, node.sourceEnd) : encodeXmlText(transformed);
+    })
+  );
+  let result = "";
+  let cursor = 0;
+  nodes.forEach((node, nodeIndex) => {
+    result += ssml.slice(cursor, node.sourceStart) + replacements[nodeIndex];
+    cursor = node.sourceEnd;
+  });
+  return result + ssml.slice(cursor);
+}
+
+// packages/ssml-core/src/azureValidation.ts
+var EXPRESS_AS_STYLES = {
+  "en-us-jennyneural": [
+    "assistant",
+    "chat",
+    "customerservice",
+    "newscast",
+    "cheerful",
+    "empathetic",
+    "excited",
+    "friendly",
+    "hopeful",
+    "sad",
+    "shouting",
+    "terrified",
+    "unfriendly",
+    "whispering"
+  ],
+  "en-us-guyneural": [
+    "angry",
+    "cheerful",
+    "excited",
+    "friendly",
+    "hopeful",
+    "newscast",
+    "sad",
+    "shouting",
+    "terrified",
+    "unfriendly",
+    "whispering"
+  ],
+  "en-us-jennymultilingualneural": [
+    "cheerful",
+    "empathetic",
+    "excited",
+    "friendly",
+    "hopeful",
+    "sad",
+    "shouting",
+    "terrified",
+    "unfriendly",
+    "whispering"
+  ],
+  "en-us-andrewneural": ["empathetic", "relieved"],
+  "ja-jp-mayuneural": ["calm", "cheerful", "sad"],
+  "ja-jp-nanamineural": ["chat", "customerservice", "cheerful", "whispering", "sad"],
+  "ja-jp-keitaneural": ["chat"],
+  "ko-kr-sunhineural": ["cheerful", "sad"],
+  "zh-cn-yunxineural": [
+    "narration-relaxed",
+    "embarrassed",
+    "fearful",
+    "sad",
+    "disgruntled",
+    "serious",
+    "angry",
+    "depressed",
+    "chat",
+    "cheerful",
+    "assistant"
+  ],
+  "zh-cn-xiaoxiaoneural": [
+    "assistant",
+    "chat",
+    "customerservice",
+    "newscast",
+    "cheerful",
+    "empathetic",
+    "excited",
+    "friendly",
+    "hopeful",
+    "sad",
+    "terrified",
+    "whispering",
+    "poetry-reading",
+    "sports_commentary",
+    "sports_commentary_excited",
+    "story"
+  ],
+  "fr-fr-deniseneural": ["cheerful", "sad"],
+  "fr-fr-henrineural": ["cheerful", "sad"],
+  "pt-br-franciscaneural": ["calm"],
+  "it-it-elsaneural": ["cheerful", "sad"],
+  "de-de-katjaneural": ["cheerful", "sad"],
+  "de-de-conradneural": ["cheerful", "sad"],
+  "ru-ru-svetlananeural": ["cheerful", "sad", "angry", "disgruntled", "embarrassed", "fearful"]
+};
+var ALLOWED_BREAK_STRENGTHS = /* @__PURE__ */ new Set(["none", "x-weak", "weak", "medium", "strong", "x-strong"]);
+var ALLOWED_SAY_AS = /* @__PURE__ */ new Set([
+  "characters",
+  "spell-out",
+  "cardinal",
+  "ordinal",
+  "number",
+  "date",
+  "time",
+  "telephone",
+  "fraction",
+  "address",
+  "name",
+  "currency"
+]);
+var ALLOWED_ROLES = /* @__PURE__ */ new Set([
+  "Girl",
+  "Boy",
+  "YoungAdultFemale",
+  "YoungAdultMale",
+  "OlderAdultFemale",
+  "OlderAdultMale",
+  "SeniorFemale",
+  "SeniorMale"
+]);
+var ALLOWED_EMPHASIS_LEVELS = /* @__PURE__ */ new Set(["strong", "moderate", "reduced", "none"]);
+var ALLOWED_SILENCE_TYPES = /* @__PURE__ */ new Set([
+  "Leading",
+  "Tailing",
+  "Sentenceboundary",
+  "Comma",
+  "Semicolon",
+  "Enumerationcomma"
+]);
+var ALLOWED_VISEME_TYPES = /* @__PURE__ */ new Set(["redlips_front", "FacialExpression"]);
+function decodeAttribute(value) {
+  return value.replace(
+    /&(?:amp|apos|gt|lt|quot);/gi,
+    (entity) => ({ "&amp;": "&", "&apos;": "'", "&gt;": ">", "&lt;": "<", "&quot;": '"' })[entity.toLowerCase()] ?? entity
+  );
+}
+function findTagEnd2(source, start) {
+  let quote = "";
+  for (let index = start; index < source.length; index += 1) {
+    const character = source[index];
+    if (quote) {
+      if (character === quote) quote = "";
+    } else if (character === '"' || character === "'") quote = character;
+    else if (character === ">") return index;
+  }
+  return source.length - 1;
+}
+function tokenizeElements(source) {
+  const tokens = [];
+  const openElements = [];
+  let index = 0;
+  while (index < source.length) {
+    const start = source.indexOf("<", index);
+    if (start === -1) break;
+    if (source.startsWith("<!--", start)) {
+      const end2 = source.indexOf("-->", start + 4);
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<![CDATA[", start)) {
+      const end2 = source.indexOf("]]>", start + 9);
+      index = end2 === -1 ? source.length : end2 + 3;
+      continue;
+    }
+    if (source.startsWith("<?", start)) {
+      const end2 = source.indexOf("?>", start + 2);
+      index = end2 === -1 ? source.length : end2 + 2;
+      continue;
+    }
+    const end = findTagEnd2(source, start + 1);
+    const raw = source.slice(start, end + 1);
+    if (raw.startsWith("</")) {
+      openElements.pop();
+      index = end + 1;
+      continue;
+    }
+    const nameMatch = /^<\s*([A-Za-z_][A-Za-z0-9_.:-]*)/.exec(raw);
+    if (!nameMatch?.[1]) {
+      index = end + 1;
+      continue;
+    }
+    const attributes = /* @__PURE__ */ new Map();
+    const attributeSource = raw.slice(nameMatch[0].length, raw.length - 1).replace(/\/\s*$/, "");
+    const attributePattern = /([A-Za-z_][A-Za-z0-9_.:-]*)\s*=\s*(["'])([\s\S]*?)\2/g;
+    for (const match of attributeSource.matchAll(attributePattern)) {
+      attributes.set(match[1].toLowerCase(), decodeAttribute(match[3]));
+    }
+    const selfClosing = /\/\s*>$/.test(raw);
+    const parentVoiceName = [...openElements].reverse().find((element) => element.voiceName)?.voiceName;
+    tokens.push({ attributes, end, name: nameMatch[1], parentVoiceName, selfClosing, start });
+    if (!selfClosing) {
+      openElements.push({
+        name: nameMatch[1],
+        voiceName: nameMatch[1].toLowerCase() === "voice" ? attributes.get("name") : parentVoiceName
+      });
+    }
+    index = end + 1;
+  }
+  return tokens;
+}
+function location(source, offset) {
+  const before = source.slice(0, Math.max(0, offset));
+  const line = before.split("\n").length;
+  return { line, column: before.length - (before.lastIndexOf("\n") + 1) + 1 };
+}
+function addDiagnostic(diagnostics, source, offset, message, severity = "error") {
+  diagnostics.push({ ...location(source, offset), message, severity });
+}
+function attr(token, name) {
+  return token.attributes.get(name.toLowerCase());
+}
+function normalizeVoiceStyleMap(customVoiceStyleMap) {
+  const map = new Map(
+    Object.entries(EXPRESS_AS_STYLES).map(([voiceName, styles]) => [voiceName.toLowerCase(), styles])
+  );
+  for (const [voiceName, styles] of Object.entries(customVoiceStyleMap ?? {})) {
+    map.set(
+      voiceName.toLowerCase(),
+      styles.map((style) => style.toLowerCase())
+    );
+  }
+  return map;
+}
+function diagnosticSeverity(policy) {
+  if (policy === "ignore") return void 0;
+  return policy === "error" ? "error" : "warning";
+}
+function validateElement(token, source, diagnostics, voiceName, options, voiceStyleMap) {
+  const name = token.name.toLowerCase();
+  if (name === "voice" && !attr(token, "name")?.trim())
+    addDiagnostic(diagnostics, source, token.start, '<voice> requires a non-empty "name" attribute.');
+  if (name === "break") {
+    const time = attr(token, "time");
+    const strength = attr(token, "strength");
+    if (!time && !strength)
+      addDiagnostic(diagnostics, source, token.start, '<break> requires either "time" or "strength".');
+    if (time && strength)
+      addDiagnostic(diagnostics, source, token.start, '<break> must not specify both "time" and "strength".');
+    if (time && !/^\d+(?:\.\d+)?(?:ms|s)$/.test(time.trim()))
+      addDiagnostic(diagnostics, source, token.start, '<break time> must use a numeric value followed by "ms" or "s".');
+    if (strength && !ALLOWED_BREAK_STRENGTHS.has(strength))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <break strength> value "${strength}".`);
+  }
+  if (name === "prosody") {
+    const rate = attr(token, "rate");
+    const pitch = attr(token, "pitch");
+    const volume = attr(token, "volume");
+    if (rate && !/^(x-slow|slow|medium|fast|x-fast|[+-]?\d+(?:\.\d+)?%)$/.test(rate.trim()))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <prosody rate> value "${rate}".`);
+    if (pitch && !/^(x-low|low|medium|high|x-high|[+-]?\d+(?:\.\d+)?(?:st|Hz|%)?)$/.test(pitch.trim()))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <prosody pitch> value "${pitch}".`);
+    if (volume && !/^(silent|x-soft|soft|medium|loud|x-loud|[+-]?\d+(?:\.\d+)?(?:dB|%)?)$/.test(volume.trim()))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <prosody volume> value "${volume}".`);
+  }
+  if (name === "mstts:express-as" || name === "express-as" || name === "expressas") {
+    const style = attr(token, "style");
+    if (!style?.trim())
+      addDiagnostic(diagnostics, source, token.start, '<mstts:express-as> requires a non-empty "style" attribute.');
+    const degree = attr(token, "styledegree") ?? attr(token, "style-degree");
+    if (degree && (!/^\d+(?:\.\d+)?$/.test(degree) || Number(degree) < 0.01 || Number(degree) > 2))
+      addDiagnostic(
+        diagnostics,
+        source,
+        token.start,
+        "<mstts:express-as styledegree> must be a number between 0.01 and 2."
+      );
+    const role = attr(token, "role");
+    if (role && !ALLOWED_ROLES.has(role))
+      addDiagnostic(diagnostics, source, token.start, `Unsupported <mstts:express-as role> value "${role}".`);
+    const supportedStyles = voiceName ? voiceStyleMap.get(voiceName.toLowerCase()) : void 0;
+    const severity = diagnosticSeverity(options.unknownVoicePolicy ?? "warn");
+    if (style && supportedStyles && !supportedStyles.includes(style.toLowerCase()) && severity)
+      addDiagnostic(
+        diagnostics,
+        source,
+        token.start,
+        `Unknown style "${style}" is not supported by voice "${voiceName}" according to the configured voice style map.`,
+        severity
+      );
+    if (style && voiceName && !supportedStyles && severity)
+      addDiagnostic(
+        diagnostics,
+        source,
+        token.start,
+        `Unknown style "${style}" cannot be verified because voice "${voiceName}" is not registered in the voice style map.`,
+        severity
+      );
+  }
+  if (name === "say-as" || name === "sayas") {
+    const interpretAs = attr(token, "interpret-as");
+    if (!interpretAs || !ALLOWED_SAY_AS.has(interpretAs))
+      addDiagnostic(diagnostics, source, token.start, `<say-as> requires a supported "interpret-as" value.`);
+  }
+  if (name === "phoneme" && (!attr(token, "alphabet") || !attr(token, "ph")))
+    addDiagnostic(diagnostics, source, token.start, '<phoneme> requires both "alphabet" and "ph" attributes.');
+  if (name === "emphasis" && attr(token, "level") && !ALLOWED_EMPHASIS_LEVELS.has(attr(token, "level") ?? ""))
+    addDiagnostic(diagnostics, source, token.start, `Unsupported <emphasis level> value "${attr(token, "level")}".`);
+  if (name === "sub" && !attr(token, "alias")?.trim())
+    addDiagnostic(diagnostics, source, token.start, '<sub> requires a non-empty "alias" attribute.');
+  if (name === "lang" && !attr(token, "xml:lang")?.trim() && !attr(token, "lang")?.trim())
+    addDiagnostic(diagnostics, source, token.start, '<lang> requires an "xml:lang" attribute.');
+  if (name === "mark" && !attr(token, "name")?.trim())
+    addDiagnostic(diagnostics, source, token.start, '<mark> requires a non-empty "name" attribute.');
+  if (name === "bookmark" && !attr(token, "mark")?.trim())
+    addDiagnostic(diagnostics, source, token.start, '<bookmark> requires a non-empty "mark" attribute.');
+  if (name === "lexicon") {
+    const uri = attr(token, "uri");
+    if (!uri) addDiagnostic(diagnostics, source, token.start, '<lexicon> requires a "uri" attribute.');
+    else {
+      try {
+        const parsed = new URL(uri);
+        if (parsed.protocol !== "https:")
+          addDiagnostic(diagnostics, source, token.start, "<lexicon uri> must use HTTPS.");
+      } catch {
+        addDiagnostic(diagnostics, source, token.start, "<lexicon uri> must be an absolute HTTPS URL.");
+      }
+    }
+  }
+  if (name === "mstts:silence") {
+    const type = attr(token, "type");
+    const value = attr(token, "value");
+    if (!type || !ALLOWED_SILENCE_TYPES.has(type))
+      addDiagnostic(diagnostics, source, token.start, '<mstts:silence> requires a supported "type" attribute.');
+    if (!value || !/^\d+(?:\.\d+)?(?:ms|s)$/.test(value.trim()))
+      addDiagnostic(diagnostics, source, token.start, '<mstts:silence> requires a time-valued "value" attribute.');
+  }
+  if (name === "mstts:viseme") {
+    const type = attr(token, "type");
+    if (!type || !ALLOWED_VISEME_TYPES.has(type))
+      addDiagnostic(diagnostics, source, token.start, '<mstts:viseme> requires a supported "type" attribute.');
+  }
+  if (name === "audio") {
+    const src = attr(token, "src");
+    if (!src) addDiagnostic(diagnostics, source, token.start, '<audio> requires a "src" attribute.');
+    else {
+      let parsed;
+      try {
+        parsed = new URL(src);
+      } catch {
+        addDiagnostic(diagnostics, source, token.start, "<audio src> must be an absolute HTTP(S) URL.");
+        return;
+      }
+      if (parsed.protocol !== "https:" && !(options.allowHttpAudio && parsed.protocol === "http:"))
+        addDiagnostic(diagnostics, source, token.start, "<audio src> must use HTTPS.");
+      if (options.allowedAudioOrigins && !options.allowedAudioOrigins.includes(parsed.origin))
+        addDiagnostic(diagnostics, source, token.start, `<audio src> origin "${parsed.origin}" is not allowed.`);
+      else if (!options.allowExternalAudio)
+        addDiagnostic(
+          diagnostics,
+          source,
+          token.start,
+          `<audio src> external origin "${parsed.origin}" is blocked by default; set allowExternalAudio to true or provide allowedAudioOrigins.`,
+          "error"
+        );
+    }
+  }
+}
+function validateAzureSsml(ssml, options = {}) {
+  const diagnostics = [];
+  if (typeof ssml !== "string") {
+    return [{ line: 1, column: 1, message: "SSML input must be a string", severity: "error" }];
+  }
+  const maxLength = options.maxLength ?? 1e4;
+  if (ssml.length > maxLength)
+    addDiagnostic(diagnostics, ssml, maxLength, `SSML exceeds the maximum length of ${maxLength} characters.`);
+  try {
+    parseSsml(ssml);
+  } catch (error) {
+    const message = error instanceof Error ? error.message.replace(/ at position \d+$/, "") : String(error);
+    const match = / at position (\d+)$/.exec(error instanceof Error ? error.message : "");
+    addDiagnostic(diagnostics, ssml, match ? Number(match[1]) : 0, message);
+    return diagnostics;
+  }
+  const tokens = tokenizeElements(ssml);
+  const speak = tokens.find((token) => token.name.toLowerCase() === "speak");
+  const voices = tokens.filter((token) => token.name.toLowerCase() === "voice");
+  if (!speak || voices.length === 0)
+    addDiagnostic(
+      diagnostics,
+      ssml,
+      speak?.start ?? 0,
+      "Azure SSML requires at least one <voice> element under <speak>."
+    );
+  const voiceName = voices[0] ? attr(voices[0], "name") : void 0;
+  const voiceStyleMap = normalizeVoiceStyleMap(options.customVoiceStyleMap);
+  const policySeverity = diagnosticSeverity(options.unknownVoicePolicy ?? "warn");
+  const voicesToValidate = options.validateNestedVoices === false ? voices.slice(0, 1) : voices;
+  for (const token of voicesToValidate) {
+    const name = attr(token, "name")?.trim();
+    if (name && !voiceStyleMap.has(name.toLowerCase()) && policySeverity)
+      addDiagnostic(
+        diagnostics,
+        ssml,
+        token.start,
+        `Unknown voice "${name}" is not registered in the voice style map.`,
+        policySeverity
+      );
+  }
+  for (const token of tokens) {
+    const tokenVoiceName = options.validateNestedVoices === false ? voiceName : token.parentVoiceName;
+    validateElement(token, ssml, diagnostics, tokenVoiceName, options, voiceStyleMap);
+  }
+  return diagnostics;
+}
+
 export {
   buildSsml,
   parseSsml,
   buildPartialSsml,
-  validateSsml
+  validateSsml,
+  extractSsmlText,
+  mapSsmlTextNodes,
+  validateAzureSsml
 };
-//# sourceMappingURL=chunk-PSN6XAFF.mjs.map
+//# sourceMappingURL=chunk-4RHHW33A.mjs.map