ssml-builder-js 2.14.0 → 2.16.0

package/dist/core.js

@@ -36,12 +36,14 @@ __export(core_exports, {
   fromPlainTextToSsml: () => fromPlainTextToSsml,
   getAzureVoiceCatalogMetadata: () => getAzureVoiceCatalogMetadata,
   getBuiltInVoiceCatalogMetadata: () => getBuiltInVoiceCatalogMetadata,
+  getSsmlSourceMap: () => getSsmlSourceMap,
   isValidAzureAudioDuration: () => isValidAzureAudioDuration,
   mapSsmlTextNodes: () => mapSsmlTextNodes,
   normalizeAzureLanguage: () => normalizeAzureLanguage,
   parseSsml: () => parseSsml,
   splitSsmlDocument: () => splitSsmlDocument,
   validateAzureSsml: () => validateAzureSsml,
+  validateAzureSsmlChunks: () => validateAzureSsmlChunks,
   validateSsml: () => validateSsml,
   validateSsmlStructureIntegrity: () => validateSsmlStructureIntegrity
 });
@@ -974,6 +976,236 @@ function buildPartialSsml(textOrOptions, context) {
   return serializePartialSsml(textOrOptions.text, textOrOptions);
 }
 
+// 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 collectSourceMap(source) {
+  const segments = [];
+  const markers = [];
+  const elements = [];
+  let textOffset = 0;
+  let index = 0;
+  const textParts = [];
+  const addText = (value) => {
+    if (!value) return;
+    const parent = elements[elements.length - 1];
+    if (parent) parent.nextChildIndex += 1;
+    const sourceNodePath = parent?.path ?? ["speak"];
+    const start = textOffset;
+    textOffset += value.length;
+    textParts.push(value);
+    segments.push({ text: value, range: { start, end: textOffset }, sourceNodePath: [...sourceNodePath] });
+  };
+  while (index < source.length) {
+    if (source[index] !== "<") {
+      const end2 = source.indexOf("<", index);
+      const textEnd = end2 === -1 ? source.length : end2;
+      addText(decodeXmlText(source.slice(index, textEnd)));
+      index = textEnd;
+      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(source.slice(contentStart, contentEnd));
+      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;
+    }
+    const end = findTagEnd(source, index + 1);
+    const rawTag = source.slice(index, end + 1);
+    if (rawTag.startsWith("</")) {
+      elements.pop();
+      index = end + 1;
+      continue;
+    }
+    const name = readTagName(rawTag);
+    if (!name) {
+      index = end + 1;
+      continue;
+    }
+    const parent = elements[elements.length - 1];
+    const childIndex = parent?.nextChildIndex ?? 0;
+    if (parent) parent.nextChildIndex += 1;
+    const path = parent ? [...parent.path, `${name}[${childIndex}]`] : [name];
+    const attributes = readTagAttributes(rawTag, name);
+    const normalizedName = name.toLowerCase();
+    if (normalizedName === "mark" || normalizedName === "bookmark") {
+      const markerName = attributes[normalizedName === "mark" ? "name" : "mark"];
+      if (markerName) {
+        markers.push({
+          kind: normalizedName,
+          name: markerName,
+          originalTextRange: { start: textOffset, end: textOffset },
+          sourceNodePath: [...path]
+        });
+      }
+    }
+    if (!/\/\s*>$/.test(rawTag)) elements.push({ name, path, nextChildIndex: 0 });
+    index = end + 1;
+  }
+  return { text: textParts.join(""), segments, markers };
+}
+function getSsmlSourceMap(ssml) {
+  parseSsml(ssml);
+  return collectSourceMap(ssml);
+}
+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/split.ts
 var DEFAULT_MAX_LENGTH = 1e4;
 function cloneElement(element, children) {
@@ -1105,7 +1337,7 @@ function findSourceNodePath(nodes, targetOffset) {
   });
   return foundPath ?? firstPath;
 }
-function createChunk(document, nodes, chunkIndex, textStart, backgroundAudio, replicateBackgroundAudio) {
+function createChunk(document, nodes, chunkIndex, textStart, backgroundAudio, replicateBackgroundAudio, sourceMap, includeEndMarkers) {
   const chunkNodes = backgroundAudio && (replicateBackgroundAudio || chunkIndex === 0) ? [backgroundAudio, ...nodes] : nodes;
   const text = nodes.map(textFromNode).join("");
   const marks = [];
@@ -1121,7 +1353,23 @@ function createChunk(document, nodes, chunkIndex, textStart, backgroundAudio, re
     hasBackgroundAudio: chunkNodes.some(
       (node) => typeof node !== "string" && node.type !== "text" && node.type === "mstts:backgroundaudio"
     ),
-    sourceNodePath: findSourceNodePath(document.children ?? [], textStart)
+    sourceNodePath: findSourceNodePath(document.children ?? [], textStart),
+    sourceTextSegments: sourceMap.segments.filter(({ range }) => range.end > textStart && range.start < textStart + text.length).map((segment) => {
+      const start = Math.max(segment.range.start, textStart);
+      const end = Math.min(segment.range.end, textStart + text.length);
+      return {
+        text: segment.text.slice(start - segment.range.start, end - segment.range.start),
+        range: { start, end },
+        sourceNodePath: [...segment.sourceNodePath]
+      };
+    }),
+    sourceMarkers: sourceMap.markers.filter(
+      ({ originalTextRange }) => originalTextRange.start >= textStart && (originalTextRange.start < textStart + text.length || includeEndMarkers && originalTextRange.start === textStart + text.length)
+    ).map((marker) => ({
+      ...marker,
+      originalTextRange: { ...marker.originalTextRange },
+      sourceNodePath: [...marker.sourceNodePath]
+    }))
   };
 }
 function splitSsmlDocument(ssml, maxLength = DEFAULT_MAX_LENGTH, options = {}) {
@@ -1131,11 +1379,12 @@ function splitSsmlDocument(ssml, maxLength = DEFAULT_MAX_LENGTH, options = {}) {
     throw new RangeError("maxLength must be a positive integer");
   }
   const document = parseSsml(ssml);
+  const sourceMap = getSsmlSourceMap(ssml);
   const backgroundAudio = (document.children ?? []).find(
     (node) => typeof node !== "string" && node.type !== "text" && node.type === "mstts:backgroundaudio"
   );
   if (ssml.length <= resolvedMaxLength) {
-    return [createChunk(document, document.children ?? [], 0, 0, backgroundAudio, true)];
+    return [createChunk(document, document.children ?? [], 0, 0, backgroundAudio, true, sourceMap, true)];
   }
   const contentChildren = (document.children ?? []).filter((node) => node !== backgroundAudio);
   const plainDocumentLength = documentWithChildren(document, []).length;
@@ -1159,7 +1408,16 @@ function splitSsmlDocument(ssml, maxLength = DEFAULT_MAX_LENGTH, options = {}) {
   }
   if (group.length > 0) chunks.push(group);
   if (chunks.length === 0) {
-    const result = createChunk(document, [], 0, 0, backgroundAudio, resolvedOptions.replicateBackgroundAudio ?? false);
+    const result = createChunk(
+      document,
+      [],
+      0,
+      0,
+      backgroundAudio,
+      resolvedOptions.replicateBackgroundAudio ?? false,
+      sourceMap,
+      true
+    );
     if (result.ssml.length > resolvedMaxLength) {
       throw new RangeError("maxLength is too small to contain the SSML document wrapper");
     }
@@ -1173,7 +1431,9 @@ function splitSsmlDocument(ssml, maxLength = DEFAULT_MAX_LENGTH, options = {}) {
       chunkIndex,
       textStart,
       backgroundAudio,
-      resolvedOptions.replicateBackgroundAudio ?? false
+      resolvedOptions.replicateBackgroundAudio ?? false,
+      sourceMap,
+      chunkIndex === chunks.length - 1
     );
     textStart = result.originalTextRange.end;
     return result;
@@ -1196,155 +1456,6 @@ 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 === "&amp;") return "&";
-    if (entity === "&apos;") return "'";
-    if (entity === "&gt;") 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/migration.ts
 var DEFAULT_TRANSLATION_SKIP_TAGS = ["phoneme", "say-as", "sayAs", "sub"];
 function elementName2(element) {
@@ -1765,34 +1876,51 @@ function createAzureUrlValidatorRunner(validator, options = {}) {
   const inFlight = /* @__PURE__ */ new Map();
   const waiters = [];
   let active = 0;
-  const acquire = async () => {
+  const configuredSignal = options.signal ?? new AbortController().signal;
+  const acquire = async (signal) => {
+    if (signal.aborted) throw new Error("URL validation was aborted.");
     if (active < concurrency) {
       active += 1;
       return;
     }
-    await new Promise((resolve) => waiters.push(resolve));
+    await new Promise((resolve, reject) => {
+      let waiter;
+      const abortHandler = () => {
+        const index = waiters.indexOf(waiter);
+        if (index >= 0) waiters.splice(index, 1);
+        signal.removeEventListener("abort", abortHandler);
+        reject(new Error("URL validation was aborted."));
+      };
+      signal.addEventListener("abort", abortHandler, { once: true });
+      waiter = () => {
+        signal.removeEventListener("abort", abortHandler);
+        resolve();
+      };
+      waiters.push(waiter);
+    });
     active += 1;
   };
   const release = () => {
     active -= 1;
     waiters.shift()?.();
   };
-  const check = async (url, context) => {
-    if (options.signal?.aborted) throw new Error("URL validation was aborted.");
-    const cached = cache.get(url);
+  const check = async (url, context, signal = configuredSignal) => {
+    if (signal.aborted) throw new Error("URL validation was aborted.");
+    const key = `${context.tag}:${context.attribute}:${url}`;
+    const cached = cache.get(key);
     if (cached !== void 0) return cached;
-    const existing = inFlight.get(url);
+    const existing = inFlight.get(key);
     if (existing) return existing;
     const promise = (async () => {
-      await acquire();
+      await acquire(signal);
       try {
-        if (options.signal?.aborted) throw new Error("URL validation was aborted.");
-        const validation = Promise.resolve(validator(url, context));
+        if (signal.aborted) throw new Error("URL validation was aborted.");
+        const validation = Promise.resolve(validator(url, context, signal));
         let timer;
         let abortHandler;
         const cancellation = new Promise((_resolve, reject) => {
           abortHandler = () => reject(new Error("URL validation was aborted."));
-          options.signal?.addEventListener("abort", abortHandler, { once: true });
+          signal.addEventListener("abort", abortHandler, { once: true });
           if (options.timeoutMs !== void 0 && options.timeoutMs > 0) {
             timer = setTimeout(
               () => reject(new Error(`URL validation timed out after ${options.timeoutMs} ms.`)),
@@ -1802,24 +1930,24 @@ function createAzureUrlValidatorRunner(validator, options = {}) {
         });
         try {
           const result = await (timer || abortHandler ? Promise.race([validation, cancellation]) : validation);
-          cache.set(url, result);
+          cache.set(key, result);
           return result;
         } finally {
           if (timer) clearTimeout(timer);
-          if (abortHandler) options.signal?.removeEventListener("abort", abortHandler);
+          if (abortHandler) signal.removeEventListener("abort", abortHandler);
         }
       } finally {
         release();
       }
     })();
-    inFlight.set(url, promise);
+    inFlight.set(key, promise);
     try {
       return await promise;
     } finally {
-      inFlight.delete(url);
+      inFlight.delete(key);
     }
   };
-  return (url, context) => check(url, context);
+  return (url, context, signal) => check(url, context, signal ?? configuredSignal);
 }
 var ALLOWED_BREAK_STRENGTHS = /* @__PURE__ */ new Set(["none", "x-weak", "weak", "medium", "strong", "x-strong"]);
 var ALLOWED_SAY_AS = /* @__PURE__ */ new Set([
@@ -1932,6 +2060,7 @@ function tokenizeElements(source) {
     if (parent) parent.childElementCount += 1;
     const parentVoiceName = [...openElements].reverse().find((element) => element.voiceName)?.voiceName;
     const tokenName = nameMatch[1];
+    const path = parent ? [...parent.path, `${tokenName}[${childElementIndex ?? 0}]`] : [tokenName];
     const tokenVoiceName = tokenName.toLowerCase() === "voice" ? attributes.get("name") : tokenName.toLowerCase() === "mstts:turn" ? attributes.get("voice") ?? parentVoiceName : parentVoiceName;
     tokens.push({
       attributes,
@@ -1942,12 +2071,14 @@ function tokenizeElements(source) {
       parentName: parent?.name,
       parentVoiceName,
       selfClosing,
-      start
+      start,
+      path
     });
     if (!selfClosing) {
       openElements.push({
         childElementCount: 0,
         name: tokenName,
+        path,
         voiceName: tokenVoiceName
       });
     }
@@ -1960,13 +2091,14 @@ function location(source, 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", code) {
+function addDiagnostic(diagnostics, source, offset, message, severity = "error", code, metadata = {}) {
   diagnostics.push({
     ...location(source, offset),
     message,
     severity,
     source: "ssml-static-validator",
-    ...code ? { code } : {}
+    ...code ? { code } : {},
+    ...metadata
   });
 }
 function isSupportedProsodyRate(value) {
@@ -2439,9 +2571,11 @@ function validateAzureSsmlStatic(ssml, options = {}) {
   for (const token of tokens) {
     const tokenName = token.name.toLowerCase();
     const tokenVoiceName = tokenName === "voice" ? attr(token, "name")?.trim() : tokenName === "mstts:turn" ? attr(token, "voice")?.trim() || token.parentVoiceName : options.validateNestedVoices === false ? voiceName : token.parentVoiceName;
+    const tokenDiagnosticStart = diagnostics.length;
     validateElement(token, ssml, diagnostics, tokenVoiceName, options, voiceCatalog);
     const definition = tokenVoiceName ? voiceCatalog.get(tokenVoiceName.toLowerCase()) : void 0;
     validateVoiceFeatureMatrix(token, ssml, diagnostics, tokenVoiceName, definition);
+    annotateTokenDiagnostics(diagnostics, tokenDiagnosticStart, token, options, tokenVoiceName);
     if (tokenName === "voice" && options.model && definition?.models && !definition.models.some((model) => model.toLowerCase() === options.model?.toLowerCase())) {
       addDiagnostic(
         diagnostics,
@@ -2463,18 +2597,37 @@ function urlAttributes(token) {
     return value === void 0 ? [] : [{ attribute, value }];
   });
 }
+function annotateTokenDiagnostics(diagnostics, startIndex, token, options, voiceName) {
+  const attributes = [...token.attributes.keys()];
+  for (const diagnostic of diagnostics.slice(startIndex)) {
+    const attributeName = attributes.find(
+      (attribute) => new RegExp(`(?:<[^> ]+\\s+|")${attribute}(?:"|>|\\s)`, "i").test(diagnostic.message)
+    );
+    const nodePath = options.sourceNodePath ? [...options.sourceNodePath] : [...token.path];
+    Object.assign(diagnostic, {
+      range: { start: token.start, end: token.end + 1 },
+      tagName: token.name,
+      ...attributeName ? { attributeName } : {},
+      ...voiceName ? { voiceName } : {},
+      ...options.chunkIndex !== void 0 ? { chunkIndex: options.chunkIndex } : {},
+      nodePath,
+      targetNodePath: [...token.path]
+    });
+  }
+}
 function validateAzureSsml(ssml, options = {}) {
   const diagnostics = validateAzureSsmlStatic(ssml, options);
   const validator = options.urlValidator ?? options.customUrlValidator;
   if (!validator || typeof ssml !== "string") return diagnostics;
   const runnerOptions = options.urlValidation ?? {};
-  const boundedValidator = createAzureUrlValidatorRunner(validator, {
+  const boundedValidator = options.urlValidatorRunner ?? createAzureUrlValidatorRunner(validator, {
     ...runnerOptions,
     ...options.urlValidatorConcurrency !== void 0 ? { concurrency: options.urlValidatorConcurrency } : {},
     ...options.urlValidatorTimeoutMs !== void 0 ? { timeoutMs: options.urlValidatorTimeoutMs } : {},
     ...options.urlValidatorSignal ? { signal: options.urlValidatorSignal } : {},
     ...options.urlValidatorCache ? { cache: options.urlValidatorCache } : {}
   });
+  const validationSignal = options.urlValidatorSignal ?? options.urlValidation?.signal ?? new AbortController().signal;
   let tokens;
   try {
     tokens = tokenizeElements(ssml);
@@ -2484,30 +2637,54 @@ function validateAzureSsml(ssml, options = {}) {
   const checks = tokens.flatMap(
     (token) => urlAttributes(token).map(async ({ attribute, value }) => {
       try {
-        const result = await boundedValidator(value, { tag: token.name, attribute });
+        const result = await boundedValidator(
+          value,
+          { tag: token.name, attribute, ...options.sourceNodePath ? { sourceNodePath: options.sourceNodePath } : {} },
+          validationSignal
+        );
         const valid = typeof result === "boolean" ? result : result.valid;
         if (!valid) {
           const reason = typeof result === "boolean" ? void 0 : result.reason;
+          const diagnosticStart = diagnostics.length;
           addDiagnostic(
             diagnostics,
             ssml,
             token.start,
             `<${token.name} ${attribute}> was rejected by the custom URL validator${reason ? `: ${reason}` : "."}`
           );
+          annotateTokenDiagnostics(diagnostics, diagnosticStart, token, options, token.parentVoiceName);
         }
       } catch (error) {
         const reason = error instanceof Error ? error.message : String(error);
+        const diagnosticStart = diagnostics.length;
         addDiagnostic(
           diagnostics,
           ssml,
           token.start,
           `<${token.name} ${attribute}> could not be validated by the custom URL validator: ${reason}`
         );
+        annotateTokenDiagnostics(diagnostics, diagnosticStart, token, options, token.parentVoiceName);
       }
     })
   );
   return Promise.all(checks).then(() => diagnostics);
 }
+async function validateAzureSsmlChunks(chunks, options = {}) {
+  const validator = options.urlValidator ?? options.customUrlValidator;
+  const sharedOptions = validator ? {
+    ...options,
+    urlValidatorRunner: options.urlValidatorRunner ?? createAzureUrlValidatorRunner(validator, {
+      ...options.urlValidation ?? {},
+      ...options.urlValidatorConcurrency !== void 0 ? { concurrency: options.urlValidatorConcurrency } : {},
+      ...options.urlValidatorTimeoutMs !== void 0 ? { timeoutMs: options.urlValidatorTimeoutMs } : {},
+      ...options.urlValidatorSignal ? { signal: options.urlValidatorSignal } : {},
+      ...options.urlValidatorCache ? { cache: options.urlValidatorCache } : {}
+    })
+  } : options;
+  return Promise.all(
+    chunks.map((chunk, chunkIndex) => Promise.resolve(validateAzureSsml(chunk, { ...sharedOptions, chunkIndex })))
+  );
+}
 
 // packages/ssml-core/src/generated/azureVoiceCatalog.ts
 var AZURE_VOICE_CATALOG_METADATA = {
@@ -2536,12 +2713,14 @@ var getBuiltInVoiceCatalogMetadata = getAzureVoiceCatalogMetadata;
   fromPlainTextToSsml,
   getAzureVoiceCatalogMetadata,
   getBuiltInVoiceCatalogMetadata,
+  getSsmlSourceMap,
   isValidAzureAudioDuration,
   mapSsmlTextNodes,
   normalizeAzureLanguage,
   parseSsml,
   splitSsmlDocument,
   validateAzureSsml,
+  validateAzureSsmlChunks,
   validateSsml,
   validateSsmlStructureIntegrity
 });