@mux/ai 0.8.2 → 0.9.0

package/dist/workflows/index.d.ts

@@ -1,4 +1,4 @@
-export { A as AskQuestionsOptions, a as AskQuestionsResult, b as AskQuestionsType, c as AudioTranslationOptions, d as AudioTranslationResult, B as BurnedInCaptionsAnalysis, e as BurnedInCaptionsOptions, f as BurnedInCaptionsPromptOverrides, g as BurnedInCaptionsPromptSections, h as BurnedInCaptionsResult, C as Chapter, j as ChapterSystemPromptSections, k as ChaptersOptions, l as ChaptersPromptOverrides, m as ChaptersPromptSections, n as ChaptersResult, o as ChaptersType, E as EmbeddingsOptions, p as EmbeddingsResult, H as HIVE_SEXUAL_CATEGORIES, q as HIVE_VIOLENCE_CATEGORIES, r as HiveModerationOutput, s as HiveModerationSource, M as ModerationOptions, t as ModerationProvider, u as ModerationResult, Q as Question, v as QuestionAnswer, w as QuestionAnswerType, S as SUMMARY_KEYWORD_LIMIT, x as SummarizationOptions, y as SummarizationPromptOverrides, z as SummarizationPromptSections, D as SummaryAndTagsResult, F as SummaryType, T as ThumbnailModerationScore, G as TranslationOptions, I as TranslationPayload, J as TranslationResult, K as askQuestions, L as burnedInCaptionsSchema, N as chapterSchema, O as chaptersSchema, P as generateChapters, R as generateEmbeddings, U as generateVideoEmbeddings, V as getModerationScores, W as getSummaryAndTags, X as hasBurnedInCaptions, Y as questionAnswerSchema, Z as summarySchema, _ as translateAudio, $ as translateCaptions, a0 as translationSchema } from '../index-DP02N3iR.js';
+export { A as AskQuestionsOptions, a as AskQuestionsResult, b as AskQuestionsType, c as AudioTranslationOptions, d as AudioTranslationResult, B as BurnedInCaptionsAnalysis, e as BurnedInCaptionsOptions, f as BurnedInCaptionsPromptOverrides, g as BurnedInCaptionsPromptSections, h as BurnedInCaptionsResult, C as Chapter, j as ChapterSystemPromptSections, k as ChaptersOptions, l as ChaptersPromptOverrides, m as ChaptersPromptSections, n as ChaptersResult, o as ChaptersType, E as EmbeddingsOptions, p as EmbeddingsResult, H as HIVE_SEXUAL_CATEGORIES, q as HIVE_VIOLENCE_CATEGORIES, r as HiveModerationOutput, s as HiveModerationSource, M as ModerationOptions, t as ModerationProvider, u as ModerationResult, Q as Question, v as QuestionAnswer, w as QuestionAnswerType, S as SUMMARY_KEYWORD_LIMIT, x as SummarizationOptions, y as SummarizationPromptOverrides, z as SummarizationPromptSections, D as SummaryAndTagsResult, F as SummaryType, T as ThumbnailModerationScore, G as TranslationOptions, I as TranslationPayload, J as TranslationResult, K as askQuestions, L as burnedInCaptionsSchema, N as chapterSchema, O as chaptersSchema, P as generateChapters, R as generateEmbeddings, U as generateVideoEmbeddings, V as getModerationScores, W as getSummaryAndTags, X as hasBurnedInCaptions, Y as questionAnswerSchema, Z as summarySchema, _ as translateAudio, $ as translateCaptions, a0 as translationSchema } from '../index-CkJStzYO.js';
 import 'zod';
 import '@ai-sdk/anthropic';
 import '@ai-sdk/google';

package/dist/workflows/index.js

@@ -837,6 +837,12 @@ function createToneSection(instruction) {
     content: instruction
   };
 }
+function createLanguageSection(languageName) {
+  return {
+    tag: "language",
+    content: `All output (title, description, keywords, chapter titles) MUST be written in ${languageName}.`
+  };
+}
 
 // src/lib/retry.ts
 var DEFAULT_RETRY_OPTIONS = {
@@ -1200,6 +1206,7 @@ var SYSTEM_PROMPT = dedent`
     - Only describe observable evidence from frames or transcript
     - Do not fabricate details or make unsupported assumptions
     - Return structured data matching the requested schema exactly
+    - Provide reasoning in the same language as the question
   </constraints>
 
   <language_guidelines>
@@ -1600,6 +1607,166 @@ async function hasBurnedInCaptions(assetId, options = {}) {
 import { generateText as generateText3, Output as Output3 } from "ai";
 import dedent3 from "dedent";
 import { z as z4 } from "zod";
+
+// src/lib/language-codes.ts
+var ISO639_1_TO_3 = {
+  // Major world languages
+  en: "eng",
+  // English
+  es: "spa",
+  // Spanish
+  fr: "fra",
+  // French
+  de: "deu",
+  // German
+  it: "ita",
+  // Italian
+  pt: "por",
+  // Portuguese
+  ru: "rus",
+  // Russian
+  zh: "zho",
+  // Chinese
+  ja: "jpn",
+  // Japanese
+  ko: "kor",
+  // Korean
+  ar: "ara",
+  // Arabic
+  hi: "hin",
+  // Hindi
+  // European languages
+  nl: "nld",
+  // Dutch
+  pl: "pol",
+  // Polish
+  sv: "swe",
+  // Swedish
+  da: "dan",
+  // Danish
+  no: "nor",
+  // Norwegian
+  fi: "fin",
+  // Finnish
+  el: "ell",
+  // Greek
+  cs: "ces",
+  // Czech
+  hu: "hun",
+  // Hungarian
+  ro: "ron",
+  // Romanian
+  bg: "bul",
+  // Bulgarian
+  hr: "hrv",
+  // Croatian
+  sk: "slk",
+  // Slovak
+  sl: "slv",
+  // Slovenian
+  uk: "ukr",
+  // Ukrainian
+  tr: "tur",
+  // Turkish
+  // Asian languages
+  th: "tha",
+  // Thai
+  vi: "vie",
+  // Vietnamese
+  id: "ind",
+  // Indonesian
+  ms: "msa",
+  // Malay
+  tl: "tgl",
+  // Tagalog/Filipino
+  // Other languages
+  he: "heb",
+  // Hebrew
+  fa: "fas",
+  // Persian/Farsi
+  bn: "ben",
+  // Bengali
+  ta: "tam",
+  // Tamil
+  te: "tel",
+  // Telugu
+  mr: "mar",
+  // Marathi
+  gu: "guj",
+  // Gujarati
+  kn: "kan",
+  // Kannada
+  ml: "mal",
+  // Malayalam
+  pa: "pan",
+  // Punjabi
+  ur: "urd",
+  // Urdu
+  sw: "swa",
+  // Swahili
+  af: "afr",
+  // Afrikaans
+  ca: "cat",
+  // Catalan
+  eu: "eus",
+  // Basque
+  gl: "glg",
+  // Galician
+  is: "isl",
+  // Icelandic
+  et: "est",
+  // Estonian
+  lv: "lav",
+  // Latvian
+  lt: "lit"
+  // Lithuanian
+};
+var ISO639_3_TO_1 = Object.fromEntries(
+  Object.entries(ISO639_1_TO_3).map(([iso1, iso3]) => [iso3, iso1])
+);
+function toISO639_3(code) {
+  const normalized = code.toLowerCase().trim();
+  if (normalized.length === 3) {
+    return normalized;
+  }
+  return ISO639_1_TO_3[normalized] ?? normalized;
+}
+function toISO639_1(code) {
+  const normalized = code.toLowerCase().trim();
+  if (normalized.length === 2) {
+    return normalized;
+  }
+  return ISO639_3_TO_1[normalized] ?? normalized;
+}
+function getLanguageCodePair(code) {
+  const normalized = code.toLowerCase().trim();
+  if (normalized.length === 2) {
+    return {
+      iso639_1: normalized,
+      iso639_3: toISO639_3(normalized)
+    };
+  } else if (normalized.length === 3) {
+    return {
+      iso639_1: toISO639_1(normalized),
+      iso639_3: normalized
+    };
+  }
+  return {
+    iso639_1: normalized,
+    iso639_3: normalized
+  };
+}
+function getLanguageName(code) {
+  const iso639_1 = toISO639_1(code);
+  try {
+    const displayNames = new Intl.DisplayNames(["en"], { type: "language" });
+    return displayNames.of(iso639_1) ?? code.toUpperCase();
+  } catch {
+    return code.toUpperCase();
+  }
+}
+
+// src/workflows/chapters.ts
 var chapterSchema = z4.object({
   startTime: z4.number(),
   title: z4.string()
@@ -1660,7 +1827,8 @@ var chapterSystemPromptBuilder = createPromptBuilder({
       content: dedent3`
         - Only use information present in the transcript
         - Return structured data that matches the requested JSON schema
-        - Do not add commentary or extra text outside the JSON`
+        - Do not add commentary or extra text outside the JSON
+        - When a <language> section is provided, all chapter titles MUST be written in that language`
     },
     qualityGuidelines: {
       tag: "quality_guidelines",
@@ -1708,7 +1876,7 @@ var chaptersPromptBuilder = createPromptBuilder({
       content: dedent3`
         - Keep titles concise and descriptive
         - Avoid filler or generic labels like "Chapter 1"
-        - Use the transcript's language and terminology`
+        - Use the transcript's terminology`
     }
   },
   sectionOrder: ["task", "outputFormat", "chapterGuidelines", "titleGuidelines"]
@@ -1717,7 +1885,8 @@ function buildUserPrompt3({
   timestampedTranscript,
   promptOverrides,
   minChaptersPerHour = 3,
-  maxChaptersPerHour = 8
+  maxChaptersPerHour = 8,
+  languageName
 }) {
   const contextSections = [
     {
@@ -1726,6 +1895,9 @@ function buildUserPrompt3({
       attributes: { format: "seconds" }
     }
   ];
+  if (languageName) {
+    contextSections.push(createLanguageSection(languageName));
+  }
   const dynamicChapterGuidelines = dedent3`
     - Create at least ${minChaptersPerHour} and at most ${maxChaptersPerHour} chapters per hour of content
     - Use start times in seconds (not HH:MM:SS)
@@ -1745,7 +1917,8 @@ async function generateChapters(assetId, languageCode, options = {}) {
     promptOverrides,
     minChaptersPerHour,
     maxChaptersPerHour,
-    credentials
+    credentials,
+    outputLanguageCode
   } = options;
   const modelConfig = resolveLanguageModelConfig({
     ...options,
@@ -1789,11 +1962,14 @@ async function generateChapters(assetId, languageCode, options = {}) {
     const contentLabel = isAudioOnly ? "transcript" : "caption track";
     throw new Error(`No usable content found in ${contentLabel}`);
   }
+  const resolvedLanguageCode = outputLanguageCode && outputLanguageCode !== "auto" ? outputLanguageCode : transcriptResult.track?.language_code ?? languageCode;
+  const languageName = resolvedLanguageCode ? getLanguageName(resolvedLanguageCode) : void 0;
   const userPrompt = buildUserPrompt3({
     timestampedTranscript,
     promptOverrides,
     minChaptersPerHour,
-    maxChaptersPerHour
+    maxChaptersPerHour,
+    languageName
   });
   let chaptersData = null;
   try {
@@ -2752,6 +2928,7 @@ var SYSTEM_PROMPT3 = dedent4`
     - Do not fabricate details or make unsupported assumptions
     - Return structured data matching the requested schema
     - Output only the JSON object; no markdown or extra text
+    - When a <language> section is provided, all output text MUST be written in that language
   </constraints>
 
   <tone_guidance>
@@ -2806,6 +2983,7 @@ var AUDIO_ONLY_SYSTEM_PROMPT = dedent4`
     - Return structured data matching the requested schema
     - Focus entirely on audio/spoken content - there are no visual elements
     - Output only the JSON object; no markdown or extra text
+    - When a <language> section is provided, all output text MUST be written in that language
   </constraints>
 
   <tone_guidance>
@@ -2836,9 +3014,13 @@ function buildUserPrompt4({
   isAudioOnly = false,
   titleLength,
   descriptionLength,
-  tagCount
+  tagCount,
+  languageName
 }) {
   const contextSections = [createToneSection(TONE_INSTRUCTIONS[tone])];
+  if (languageName) {
+    contextSections.push(createLanguageSection(languageName));
+  }
   if (transcriptText) {
     const format = isCleanTranscript ? "plain text" : "WebVTT";
     contextSections.push(createTranscriptSection(transcriptText, format));
@@ -2951,7 +3133,8 @@ async function getSummaryAndTags(assetId, options) {
     credentials,
     titleLength,
     descriptionLength,
-    tagCount
+    tagCount,
+    outputLanguageCode
   } = options ?? {};
   if (!VALID_TONES.includes(tone)) {
     throw new Error(
@@ -2978,12 +3161,15 @@ async function getSummaryAndTags(assetId, options) {
       "Signed playback ID requires signing credentials. Set MUX_SIGNING_KEY and MUX_PRIVATE_KEY environment variables."
     );
   }
-  const transcriptText = includeTranscript ? (await fetchTranscriptForAsset(assetData, playbackId, {
+  const transcriptResult = includeTranscript ? await fetchTranscriptForAsset(assetData, playbackId, {
     cleanTranscript,
     shouldSign: policy === "signed",
     credentials: workflowCredentials,
     required: isAudioOnly
-  })).transcriptText : "";
+  }) : void 0;
+  const transcriptText = transcriptResult?.transcriptText ?? "";
+  const resolvedLanguageCode = outputLanguageCode && outputLanguageCode !== "auto" ? outputLanguageCode : transcriptResult?.track?.language_code ?? getReadyTextTracks(assetData)[0]?.language_code;
+  const languageName = resolvedLanguageCode ? getLanguageName(resolvedLanguageCode) : void 0;
   const userPrompt = buildUserPrompt4({
     tone,
     transcriptText,
@@ -2992,7 +3178,8 @@ async function getSummaryAndTags(assetId, options) {
     isAudioOnly,
     titleLength,
     descriptionLength,
-    tagCount
+    tagCount,
+    languageName
   });
   let analysisResponse;
   let imageUrl;
@@ -3062,164 +3249,6 @@ async function getSummaryAndTags(assetId, options) {
   };
 }
 
-// src/lib/language-codes.ts
-var ISO639_1_TO_3 = {
-  // Major world languages
-  en: "eng",
-  // English
-  es: "spa",
-  // Spanish
-  fr: "fra",
-  // French
-  de: "deu",
-  // German
-  it: "ita",
-  // Italian
-  pt: "por",
-  // Portuguese
-  ru: "rus",
-  // Russian
-  zh: "zho",
-  // Chinese
-  ja: "jpn",
-  // Japanese
-  ko: "kor",
-  // Korean
-  ar: "ara",
-  // Arabic
-  hi: "hin",
-  // Hindi
-  // European languages
-  nl: "nld",
-  // Dutch
-  pl: "pol",
-  // Polish
-  sv: "swe",
-  // Swedish
-  da: "dan",
-  // Danish
-  no: "nor",
-  // Norwegian
-  fi: "fin",
-  // Finnish
-  el: "ell",
-  // Greek
-  cs: "ces",
-  // Czech
-  hu: "hun",
-  // Hungarian
-  ro: "ron",
-  // Romanian
-  bg: "bul",
-  // Bulgarian
-  hr: "hrv",
-  // Croatian
-  sk: "slk",
-  // Slovak
-  sl: "slv",
-  // Slovenian
-  uk: "ukr",
-  // Ukrainian
-  tr: "tur",
-  // Turkish
-  // Asian languages
-  th: "tha",
-  // Thai
-  vi: "vie",
-  // Vietnamese
-  id: "ind",
-  // Indonesian
-  ms: "msa",
-  // Malay
-  tl: "tgl",
-  // Tagalog/Filipino
-  // Other languages
-  he: "heb",
-  // Hebrew
-  fa: "fas",
-  // Persian/Farsi
-  bn: "ben",
-  // Bengali
-  ta: "tam",
-  // Tamil
-  te: "tel",
-  // Telugu
-  mr: "mar",
-  // Marathi
-  gu: "guj",
-  // Gujarati
-  kn: "kan",
-  // Kannada
-  ml: "mal",
-  // Malayalam
-  pa: "pan",
-  // Punjabi
-  ur: "urd",
-  // Urdu
-  sw: "swa",
-  // Swahili
-  af: "afr",
-  // Afrikaans
-  ca: "cat",
-  // Catalan
-  eu: "eus",
-  // Basque
-  gl: "glg",
-  // Galician
-  is: "isl",
-  // Icelandic
-  et: "est",
-  // Estonian
-  lv: "lav",
-  // Latvian
-  lt: "lit"
-  // Lithuanian
-};
-var ISO639_3_TO_1 = Object.fromEntries(
-  Object.entries(ISO639_1_TO_3).map(([iso1, iso3]) => [iso3, iso1])
-);
-function toISO639_3(code) {
-  const normalized = code.toLowerCase().trim();
-  if (normalized.length === 3) {
-    return normalized;
-  }
-  return ISO639_1_TO_3[normalized] ?? normalized;
-}
-function toISO639_1(code) {
-  const normalized = code.toLowerCase().trim();
-  if (normalized.length === 2) {
-    return normalized;
-  }
-  return ISO639_3_TO_1[normalized] ?? normalized;
-}
-function getLanguageCodePair(code) {
-  const normalized = code.toLowerCase().trim();
-  if (normalized.length === 2) {
-    return {
-      iso639_1: normalized,
-      iso639_3: toISO639_3(normalized)
-    };
-  } else if (normalized.length === 3) {
-    return {
-      iso639_1: toISO639_1(normalized),
-      iso639_3: normalized
-    };
-  }
-  return {
-    iso639_1: normalized,
-    iso639_3: normalized
-  };
-}
-function getLanguageName(code) {
-  const iso639_1 = toISO639_1(code);
-  try {
-    const displayNames = new Intl.DisplayNames(["en"], { type: "language" });
-    return displayNames.of(iso639_1) ?? code.toUpperCase();
-  } catch {
-    return code.toUpperCase();
-  }
-}
-
 // src/lib/s3-sigv4.ts
 var AWS4_ALGORITHM = "AWS4-HMAC-SHA256";
 var AWS4_REQUEST_TERMINATOR = "aws4_request";