@mux/ai 0.23.0 → 0.24.0

package/dist/workflows/index.d.ts

@@ -1,4 +1,4 @@
-export { A as AskQuestionsOptions, j as AskQuestionsResult, k as AskQuestionsType, l as AudioTranslationOptions, m as AudioTranslationResult, n as AutoCensorProfanityOptions, B as BurnedInCaptionsAnalysis, o as BurnedInCaptionsOptions, p as BurnedInCaptionsPromptOverrides, q as BurnedInCaptionsPromptSections, r as BurnedInCaptionsResult, C as CaptionReplacement, s as CensorMode, u as Chapter, v as ChapterSystemPromptSections, w as ChaptersOptions, x as ChaptersPromptOverrides, y as ChaptersPromptSections, z as ChaptersResult, D as ChaptersType, E as DEFAULT_DESCRIPTION_LENGTH, F as DEFAULT_SUMMARY_KEYWORD_LIMIT, G as DEFAULT_TITLE_LENGTH, H as EditCaptionsOptions, J as EditCaptionsResult, K as EmbeddingsOptions, M as EmbeddingsResult, N as EngagementInsightsOptions, O as EngagementInsightsResult, P as GOOGLE_VISION_LIKELIHOOD_TO_SCORE, Q as HIVE_SEXUAL_CATEGORIES, R as HIVE_VIOLENCE_CATEGORIES, T as HiveModerationOutput, U as HiveModerationSource, V as ModerationOptions, W as ModerationProvider, X as ModerationResult, Y as MomentInsight, Z as OverallInsight, _ as ProfanityDetectionPayload, $ as Question, a0 as QuestionAnswer, a1 as QuestionAnswerType, a2 as ReplacementRecord, a3 as SummarizationOptions, a4 as SummarizationPromptOverrides, a5 as SummarizationPromptSections, a6 as SummaryAndTagsResult, a7 as SummaryType, a8 as ThumbnailModerationScore, a9 as TranslationChunkingOptions, aa as TranslationOptions, ab as TranslationPayload, ac as TranslationResult, ad as aggregateTokenUsage, ae as applyOverrideLists, af as applyReplacements, ag as askQuestions, ah as buildReplacementRegex, ai as burnedInCaptionsSchema, aj as censorVttContent, ak as chapterSchema, al as chaptersSchema, am as createReplacer, an as editCaptions, ao as extractTranscriptSegmentsForHotspots, ap as generateChapters, aq as generateEmbeddings, ar as generateEngagementInsights, as as generateVideoEmbeddings, at as getModerationScores, au as getSummaryAndTags, av as hasBurnedInCaptions, aw as mapShotsToHotspots, ax as normalizeTranslatedVtt, ay as profanityDetectionSchema, az as questionAnswerSchema, aA as shouldSplitChunkTranslationError, aB as summarySchema, aC as transformCueText, aD as translateAudio, aE as translateCaptions, aF as translationSchema } from '../index-ePt2EMpB.js';
+export { A as AskQuestionsOptions, j as AskQuestionsResult, k as AskQuestionsType, l as AudioTranslationOptions, m as AudioTranslationResult, n as AutoCensorProfanityOptions, B as BurnedInCaptionsAnalysis, o as BurnedInCaptionsOptions, p as BurnedInCaptionsPromptOverrides, q as BurnedInCaptionsPromptSections, r as BurnedInCaptionsResult, C as CaptionReplacement, s as CensorMode, u as Chapter, v as ChapterSystemPromptSections, w as ChaptersOptions, x as ChaptersPromptOverrides, y as ChaptersPromptSections, z as ChaptersResult, D as ChaptersType, E as DEFAULT_DESCRIPTION_LENGTH, F as DEFAULT_SUMMARY_KEYWORD_LIMIT, G as DEFAULT_TITLE_LENGTH, H as EditCaptionsOptions, J as EditCaptionsResult, K as EmbeddingsOptions, M as EmbeddingsResult, N as EngagementInsightsOptions, O as EngagementInsightsResult, P as GOOGLE_VISION_LIKELIHOOD_TO_SCORE, Q as HIVE_SEXUAL_CATEGORIES, R as HIVE_VIOLENCE_CATEGORIES, T as HiveModerationOutput, U as HiveModerationSource, V as ModerationOptions, W as ModerationProvider, X as ModerationResult, Y as MomentInsight, Z as OverallInsight, _ as ProfanityDetectionPayload, $ as Question, a0 as QuestionAnswer, a1 as QuestionAnswerType, a2 as ReplacementRecord, a3 as SummarizationOptions, a4 as SummarizationPromptOverrides, a5 as SummarizationPromptSections, a6 as SummaryAndTagsResult, a7 as SummaryType, a8 as ThumbnailModerationScore, a9 as TranslationChunkingOptions, aa as TranslationOptions, ab as TranslationPayload, ac as TranslationResult, ad as aggregateTokenUsage, ae as applyOverrideLists, af as applyReplacements, ag as askQuestions, ah as buildReplacementRegex, ai as burnedInCaptionsSchema, aj as censorVttContent, ak as chapterSchema, al as chaptersSchema, am as createReplacer, an as editCaptions, ao as extractTranscriptSegmentsForHotspots, ap as generateChapters, aq as generateEmbeddings, ar as generateEngagementInsights, as as generateVideoEmbeddings, at as getModerationScores, au as getSummaryAndTags, av as hasBurnedInCaptions, aw as mapShotsToHotspots, ax as normalizeTranslatedVtt, ay as profanityDetectionSchema, az as questionAnswerSchema, aA as shouldSplitChunkTranslationError, aB as summarySchema, aC as transformCueText, aD as translateAudio, aE as translateCaptions, aF as translationSchema } from '../index-C5zhbQHK.js';
 import 'zod';
 import '@ai-sdk/anthropic';
 import '@ai-sdk/google';

package/dist/workflows/index.js

@@ -1508,7 +1508,8 @@ function getReliableLanguageCode(track, confidenceThreshold = LOW_CONFIDENCE_THR
     return void 0;
   if (isUndeterminedLanguageCode(track.language_code))
     return void 0;
-  if (track.auto_language_confidence !== void 0 && track.auto_language_confidence < confidenceThreshold) {
+  const { auto_language_confidence: autoLanguageConfidence } = track;
+  if (autoLanguageConfidence !== void 0 && autoLanguageConfidence < confidenceThreshold) {
     return void 0;
   }
   return track.language_code;
@@ -3396,12 +3397,64 @@ function transformCueText(rawVtt, transform) {
   });
   return transformed.join("\n");
 }
+function transformCueTextBlocks(rawVtt, transform) {
+  const lines = rawVtt.split("\n");
+  const transformed = [];
+  let cueTextLines;
+  let currentCueStartTime = 0;
+  const flushCueText = () => {
+    if (!cueTextLines)
+      return;
+    if (cueTextLines.length > 0) {
+      transformed.push(...transform(cueTextLines.join("\n"), currentCueStartTime).split("\n"));
+    }
+    cueTextLines = void 0;
+  };
+  for (const line of lines) {
+    if (line.includes("-->")) {
+      flushCueText();
+      const startTimestamp = line.split("-->")[0].trim();
+      currentCueStartTime = vttTimestampToSeconds(startTimestamp);
+      transformed.push(line);
+      cueTextLines = [];
+      continue;
+    }
+    if (cueTextLines) {
+      if (line.trim() === "") {
+        flushCueText();
+        transformed.push(line);
+      } else {
+        cueTextLines.push(line);
+      }
+      continue;
+    }
+    transformed.push(line);
+  }
+  flushCueText();
+  return transformed.join("\n");
+}
+function escapeRegex(text) {
+  return text.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+}
+function buildStaticReplacementRegex(find, caseSensitive) {
+  const pattern = find.split(/(\s+)/).map((part) => /\s+/.test(part) ? "\\s+" : escapeRegex(part)).join("");
+  const firstNonWhitespaceIndex = find.search(/\S/);
+  const firstNonWhitespace = firstNonWhitespaceIndex >= 0 ? find[firstNonWhitespaceIndex] : void 0;
+  const lastNonWhitespaceMatch = /\S(?=\s*$)/.exec(find);
+  const lastNonWhitespace = lastNonWhitespaceMatch?.[0];
+  const startsAtTextBoundary = firstNonWhitespaceIndex === 0 && !!firstNonWhitespace && /\w/.test(firstNonWhitespace);
+  const endsAtTextBoundary = !!lastNonWhitespace && find.trimEnd().length === find.length && /\w/.test(lastNonWhitespace);
+  const prefix = startsAtTextBoundary ? "(?<!\\w)" : "";
+  const suffix = endsAtTextBoundary ? "(?!\\w)" : "";
+  const flags = caseSensitive === false ? "gi" : "g";
+  return new RegExp(`${prefix}${pattern}${suffix}`, flags);
+}
 function buildReplacementRegex(words) {
   const filtered = words.filter((w) => w.length > 0);
   if (filtered.length === 0)
     return null;
   filtered.sort((a, b) => b.length - a.length);
-  const escaped = filtered.map((w) => w.replace(/[.*+?^${}()|[\]\\]/g, "\\$&"));
+  const escaped = filtered.map(escapeRegex);
   const pattern = escaped.join("|");
   return new RegExp(`\\b(?:${pattern})\\b`, "gi");
 }
@@ -3453,12 +3506,10 @@ function applyReplacements(rawVtt, replacements) {
     return { editedVtt: rawVtt, replacements: [] };
   }
   const records = [];
-  const editedVtt = transformCueText(rawVtt, (line, cueStartTime) => {
-    let result = line;
+  const editedVtt = transformCueTextBlocks(rawVtt, (cueText, cueStartTime) => {
+    let result = cueText;
     for (const { find, replace, caseSensitive } of filtered) {
-      const escaped = find.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
-      const flags = caseSensitive === false ? "gi" : "g";
-      const regex = new RegExp(`\\b${escaped}\\b`, flags);
+      const regex = buildStaticReplacementRegex(find, caseSensitive);
       result = result.replace(regex, (match) => {
         records.push({ cueStartTime, before: match, after: replace });
         return replace;