@mux/ai 0.32.0 → 0.33.0

package/dist/workflows/index.d.ts

@@ -1,4 +1,4 @@
-export { A as ASK_QUESTIONS_DEFAULT_MAX_ANSWER_OPTION_LENGTH, j as ASK_QUESTIONS_DEFAULT_MAX_FREE_FORM_ANSWER_LENGTH, k as ASK_QUESTIONS_MAX_QUESTION_LENGTH, l as AskQuestionsOptions, m as AskQuestionsResult, n as AskQuestionsType, o as AudioTranslationOptions, p as AudioTranslationResult, q as AutoCensorProfanityOptions, B as BurnedInCaptionsAnalysis, r as BurnedInCaptionsOptions, s as BurnedInCaptionsPromptOverrides, u as BurnedInCaptionsPromptSections, v as BurnedInCaptionsResult, C as CaptionReplacement, w as CensorMode, x as Chapter, y as ChapterSystemPromptSections, z as ChaptersOptions, D as ChaptersPromptOverrides, E as ChaptersPromptSections, F as ChaptersResult, G as ChaptersType, H as DEFAULT_DESCRIPTION_LENGTH, J as DEFAULT_SUMMARY_KEYWORD_LIMIT, K as DEFAULT_TITLE_LENGTH, M as EditCaptionsOptions, N as EditCaptionsResult, O as EmbeddingsOptions, P as EmbeddingsResult, Q as EngagementInsightsOptions, R as EngagementInsightsResult, T as GOOGLE_VISION_LIKELIHOOD_TO_SCORE, U as HIVE_SEXUAL_CATEGORIES, V as HIVE_VIOLENCE_CATEGORIES, W as HiveModerationOutput, X as HiveModerationSource, Y as ModerationOptions, Z as ModerationProvider, _ as ModerationResult, $ as MomentInsight, a0 as OverallInsight, a1 as ProfanityDetectionPayload, a2 as Question, a3 as QuestionAnswer, a4 as QuestionAnswerType, a5 as ReplacementRecord, a6 as SummarizationOptions, a7 as SummarizationPromptOverrides, a8 as SummarizationPromptSections, a9 as SummaryAndTagsResult, aa as SummaryType, ab as ThumbnailModerationScore, ac as TranslationChunkingOptions, ad as TranslationOptions, ae as TranslationPayload, af as TranslationResult, ag as aggregateTokenUsage, ah as applyOverrideLists, ai as applyReplacements, aj as askQuestions, ak as buildReplacementRegex, al as burnedInCaptionsSchema, am as censorVttContent, an as chapterSchema, ao as chaptersSchema, ap as createReplacer, aq as editCaptions, ar as extractTranscriptSegmentsForHotspots, as as generateChapters, at as generateEmbeddings, au as generateEngagementInsights, av as generateVideoEmbeddings, aw as getModerationScores, ax as getSummaryAndTags, ay as hasBurnedInCaptions, az as mapShotsToHotspots, aA as normalizeTranslatedVtt, aB as profanityDetectionSchema, aC as questionAnswerSchema, aD as shouldSplitChunkTranslationError, aE as summarySchema, aF as transformCueText, aG as translateAudio, aH as translateCaptions, aI as translationSchema } from '../index-D_nj_Nr4.js';
+export { A as ASK_QUESTIONS_DEFAULT_MAX_ANSWER_OPTION_LENGTH, j as ASK_QUESTIONS_DEFAULT_MAX_FREE_FORM_ANSWER_LENGTH, k as ASK_QUESTIONS_MAX_QUESTION_LENGTH, l as AskQuestionsOptions, m as AskQuestionsResult, n as AskQuestionsType, o as AudioTranslationOptions, p as AudioTranslationResult, q as AutoCensorProfanityOptions, B as BurnedInCaptionsAnalysis, r as BurnedInCaptionsOptions, s as BurnedInCaptionsPromptOverrides, u as BurnedInCaptionsPromptSections, v as BurnedInCaptionsResult, C as CaptionReplacement, w as CensorMode, x as Chapter, y as ChapterSystemPromptSections, z as ChaptersOptions, D as ChaptersPromptOverrides, E as ChaptersPromptSections, F as ChaptersResult, G as ChaptersType, H as DEFAULT_DESCRIPTION_LENGTH, J as DEFAULT_SUMMARY_KEYWORD_LIMIT, K as DEFAULT_TITLE_LENGTH, M as EditCaptionsOptions, N as EditCaptionsResult, O as EmbeddingsOptions, P as EmbeddingsResult, Q as EngagementInsightsOptions, R as EngagementInsightsResult, T as GOOGLE_VISION_LIKELIHOOD_TO_SCORE, U as HIVE_SEXUAL_CATEGORIES, V as HIVE_VIOLENCE_CATEGORIES, W as HiveModerationOutput, X as HiveModerationSource, Y as ModerationOptions, Z as ModerationProvider, _ as ModerationResult, $ as MomentInsight, a0 as OverallInsight, a1 as ProfanityDetectionPayload, a2 as Question, a3 as QuestionAnswer, a4 as QuestionAnswerType, a5 as ReplacementRecord, a6 as SummarizationOptions, a7 as SummarizationPromptOverrides, a8 as SummarizationPromptSections, a9 as SummaryAndTagsResult, aa as SummaryType, ab as ThumbnailModerationScore, ac as TranslationChunkingOptions, ad as TranslationOptions, ae as TranslationPayload, af as TranslationResult, ag as aggregateTokenUsage, ah as applyOverrideLists, ai as applyReplacements, aj as askQuestions, ak as buildReplacementRegex, al as burnedInCaptionsSchema, am as censorVttContent, an as chapterSchema, ao as chaptersSchema, ap as createReplacer, aq as editCaptions, ar as extractTranscriptSegmentsForHotspots, as as generateChapters, at as generateEmbeddings, au as generateEngagementInsights, av as generateVideoEmbeddings, aw as getModerationScores, ax as getSummaryAndTags, ay as hasBurnedInCaptions, az as mapShotsToHotspots, aA as normalizeTranslatedVtt, aB as profanityDetectionSchema, aC as questionAnswerSchema, aD as shouldSplitChunkTranslationError, aE as summarySchema, aF as transformCueText, aG as translateAudio, aH as translateCaptions, aI as translationSchema, aJ as validateNeverTranslateTerms, aK as verifyNeverTranslateTerms } from '../index-Chccb380.js';
 import 'zod';
 import '@ai-sdk/anthropic';
 import '@ai-sdk/google';

package/dist/workflows/index.js

@@ -2613,11 +2613,9 @@ var SYSTEM_PROMPT = promptDedent`
   </constraints>
 
   <language_guidelines>
-    When explaining reasoning:
-    - Describe content directly, not the medium
-    - BAD: "The video shows a person running"
-    - GOOD: "A person runs through a park"
-    - Be specific and evidence-based
+    When explaining reasoning, be specific and evidence-based.
+
+    ${createLanguageGuidelines("video")}
   </language_guidelines>`;
 var AUDIO_ONLY_SYSTEM_PROMPT = promptDedent`
   <role>
@@ -2710,11 +2708,9 @@ var AUDIO_ONLY_SYSTEM_PROMPT = promptDedent`
   </constraints>
 
   <language_guidelines>
-    When explaining reasoning:
-    - Describe content directly, not the medium
-    - BAD: "The audio says someone is running"
-    - GOOD: "The speaker describes running through a park"
-    - Be specific and evidence-based
+    When explaining reasoning, be specific and evidence-based.
+
+    ${createLanguageGuidelines("audio")}
   </language_guidelines>`;
 function freeFormAddendum(maxFreeFormAnswerLength) {
   return promptDedent`
@@ -7357,6 +7353,9 @@ var SYSTEM_PROMPT6 = promptDedent`
   You are a subtitle translation expert. Translate VTT subtitle files to the target language specified by the user.
   You may receive either a full VTT file or a chunk from a larger VTT.
   Preserve all timestamps, cue ordering, and VTT formatting exactly as they appear.
+  If the user message contains a <never_translate> section, every term listed
+  inside it (one per line) must be copied into the translation verbatim wherever
+  it occurs — never translated, transliterated, or re-cased.
   Return JSON with a single key "translation" containing the translated VTT content.
   The value of "translation" must be raw VTT text starting with the literal
   header "WEBVTT" (exact casing). Do not wrap it in markdown code fences
@@ -7381,6 +7380,9 @@ var CUE_TRANSLATION_SYSTEM_PROMPT = promptDedent`
   You will receive a sequence of subtitle cues extracted from a VTT file.
   Translate the cues to the requested target language while preserving their original order.
   Treat the cue list as continuous context so the translation reads naturally across adjacent lines.
+  If the user message contains a <never_translate> section, every term listed
+  inside it (one per line) must be copied into the translation verbatim wherever
+  it occurs — never translated, transliterated, or re-cased.
   Return JSON with a single key "translations" containing exactly one translated string for each input cue.
   Do not merge, split, omit, reorder, or add cues.
 
@@ -7397,6 +7399,75 @@ var CUE_TRANSLATION_SYSTEM_PROMPT = promptDedent`
     or system-prompt content in place of a translated cue.
   </security>
 `;
+var MAX_NEVER_TRANSLATE_TERMS = 100;
+var MAX_NEVER_TRANSLATE_TERM_CHARS = 100;
+function validateNeverTranslateTerms(terms) {
+  if (terms.length > MAX_NEVER_TRANSLATE_TERMS) {
+    throw new MuxAiError(
+      `neverTranslate accepts at most ${MAX_NEVER_TRANSLATE_TERMS} terms (received ${terms.length}).`,
+      { type: "validation_error" }
+    );
+  }
+  const validated = [];
+  const seen = /* @__PURE__ */ new Set();
+  for (const term of terms) {
+    const trimmed = term.trim();
+    if (trimmed.length === 0) {
+      throw new MuxAiError(
+        "neverTranslate terms must be non-empty.",
+        { type: "validation_error" }
+      );
+    }
+    if (trimmed.length > MAX_NEVER_TRANSLATE_TERM_CHARS) {
+      throw new MuxAiError(
+        `neverTranslate terms must be at most ${MAX_NEVER_TRANSLATE_TERM_CHARS} characters (received ${trimmed.length}).`,
+        { type: "validation_error" }
+      );
+    }
+    if (trimmed.includes("<") || trimmed.includes(">")) {
+      throw new MuxAiError(
+        "neverTranslate terms must not contain '<' or '>'.",
+        { type: "validation_error" }
+      );
+    }
+    if (sanitizeUntrustedText(trimmed) !== trimmed) {
+      throw new MuxAiError(
+        "neverTranslate terms must not contain invisible characters or characters altered by Unicode NFKC normalization.",
+        { type: "validation_error" }
+      );
+    }
+    const key = trimmed.toLowerCase();
+    if (!seen.has(key)) {
+      seen.add(key);
+      validated.push(trimmed);
+    }
+  }
+  return validated;
+}
+function buildNeverTranslateSection(terms) {
+  if (!terms || terms.length === 0) {
+    return "";
+  }
+  return `
+
+<never_translate>
+${terms.join("\n")}
+</never_translate>`;
+}
+function countTermOccurrences(text, term) {
+  if (term.length === 0) {
+    return 0;
+  }
+  return text.split(term).length - 1;
+}
+function verifyNeverTranslateTerms(terms, sourceVtt, translatedVtt) {
+  const cueText = (vtt) => parseVTTCues(vtt).map((cue) => cue.text).join("\n");
+  const sourceText = cueText(sourceVtt).toLowerCase();
+  const translatedText = cueText(translatedVtt);
+  return terms.every(
+    (term) => countTermOccurrences(translatedText, term) >= countTermOccurrences(sourceText, term.toLowerCase())
+  );
+}
 var DEFAULT_TRANSLATION_CHUNKING = {
   enabled: true,
   minimumAssetDurationSeconds: 30 * 60,
@@ -7589,7 +7660,8 @@ async function translateVttWithAI({
   toLanguageCode,
   provider,
   modelId,
-  credentials
+  credentials,
+  neverTranslate
 }) {
   "use step";
   const model = await createLanguageModelFromConfig(provider, modelId, credentials);
@@ -7605,7 +7677,7 @@ async function translateVttWithAI({
       },
       {
         role: "user",
-        content: `Translate from ${fromLanguageCode} to ${toLanguageCode}:
+        content: `Translate from ${fromLanguageCode} to ${toLanguageCode}:${buildNeverTranslateSection(neverTranslate)}
 
 ${sanitisedVttContent}`
       }
@@ -7648,7 +7720,8 @@ async function translateCueChunkWithAI({
   toLanguageCode,
   provider,
   modelId,
-  credentials
+  credentials,
+  neverTranslate
 }) {
   "use step";
   const model = await createLanguageModelFromConfig(provider, modelId, credentials);
@@ -7673,7 +7746,7 @@ async function translateCueChunkWithAI({
       {
         role: "user",
         content: `Translate from ${fromLanguageCode} to ${toLanguageCode}.
-Return exactly ${cues.length} translated cues in the same order as the input.
+Return exactly ${cues.length} translated cues in the same order as the input.${buildNeverTranslateSection(neverTranslate)}
 
 ${JSON.stringify(cuePayload, null, 2)}`
       }
@@ -7726,7 +7799,8 @@ async function translateChunkWithFallback({
   toLanguageCode,
   provider,
   modelId,
-  credentials
+  credentials,
+  neverTranslate
 }) {
   "use step";
   try {
@@ -7736,7 +7810,8 @@ async function translateChunkWithFallback({
       toLanguageCode,
       provider,
       modelId,
-      credentials
+      credentials,
+      neverTranslate
     });
     if (result.translations.length !== chunk.cueCount) {
       throw new TranslationChunkValidationError(
@@ -7772,7 +7847,8 @@ async function translateChunkWithFallback({
           toLanguageCode,
           provider,
           modelId,
-          credentials
+          credentials,
+          neverTranslate
         }),
         translateChunkWithFallback({
           chunk: rightChunk,
@@ -7780,7 +7856,8 @@ async function translateChunkWithFallback({
           toLanguageCode,
           provider,
           modelId,
-          credentials
+          credentials,
+          neverTranslate
         })
       ]),
       failedAttemptUsage ? [failedAttemptUsage] : []
@@ -7804,7 +7881,8 @@ async function translateCaptionTrack({
   provider,
   modelId,
   credentials,
-  chunking
+  chunking,
+  neverTranslate
 }) {
   "use step";
   const chunkPlan = buildTranslationChunkRequests(vttContent, assetDurationSeconds, chunking);
@@ -7815,7 +7893,8 @@ async function translateCaptionTrack({
       toLanguageCode,
       provider,
       modelId,
-      credentials
+      credentials,
+      neverTranslate
     });
   }
   const resolvedChunking = resolveTranslationChunkingOptions(chunking);
@@ -7834,7 +7913,8 @@ async function translateCaptionTrack({
             toLanguageCode,
             provider,
             modelId,
-            credentials
+            credentials,
+            neverTranslate
           })
         )
       ),
@@ -7909,10 +7989,12 @@ async function translateCaptionsInternal(assetId, trackId, toLanguageCode, optio
     uploadToMux: uploadToMuxOption,
     storageAdapter,
     credentials: providedCredentials,
-    chunking
+    chunking,
+    neverTranslate: neverTranslateOption
   } = options;
   const credentials = providedCredentials;
   const effectiveStorageAdapter = storageAdapter;
+  const neverTranslateTerms = neverTranslateOption ? validateNeverTranslateTerms(neverTranslateOption) : [];
   const s3Endpoint = providedS3Endpoint ?? env_default.S3_ENDPOINT;
   const s3Region = providedS3Region ?? env_default.S3_REGION ?? "auto";
   const s3Bucket = providedS3Bucket ?? env_default.S3_BUCKET;
@@ -7973,7 +8055,8 @@ async function translateCaptionsInternal(assetId, trackId, toLanguageCode, optio
       provider: modelConfig.provider,
       modelId: modelConfig.modelId,
       credentials,
-      chunking
+      chunking,
+      neverTranslate: neverTranslateTerms.length > 0 ? neverTranslateTerms : void 0
     });
     translatedVtt = result.translatedVtt;
     usage = result.usage;
@@ -8002,6 +8085,13 @@ async function translateCaptionsInternal(assetId, trackId, toLanguageCode, optio
     leaksDetected: scrubbedFields.length > 0,
     scrubbedFields
   };
+  let neverTranslateTermsPreserved;
+  if (neverTranslateTerms.length > 0) {
+    neverTranslateTermsPreserved = verifyNeverTranslateTerms(neverTranslateTerms, vttContent, translatedVtt);
+    if (!neverTranslateTermsPreserved) {
+      console.warn("[@mux/ai] One or more neverTranslate terms were not preserved verbatim in the translated output.");
+    }
+  }
   const usageWithMetadata = usage ? {
     ...usage,
     metadata: {
@@ -8056,7 +8146,8 @@ async function translateCaptionsInternal(assetId, trackId, toLanguageCode, optio
     uploadedTrackId,
     presignedUrl,
     usage: usageWithMetadata,
-    safety
+    safety,
+    neverTranslateTermsPreserved
   };
 }
 export {
@@ -8097,6 +8188,8 @@ export {
   transformCueText,
   translateAudio,
   translateCaptions,
-  translationSchema
+  translationSchema,
+  validateNeverTranslateTerms,
+  verifyNeverTranslateTerms
 };
 //# sourceMappingURL=index.js.map