@mux/ai 0.35.0 → 0.37.0

package/dist/{shots-BMbViFbZ.d.ts → shots-CUI09Ohe.d.ts}

@@ -89,6 +89,12 @@ interface WorkflowMuxClient {
 interface MuxAIOptions {
     /** Optional timeout (ms) for helper utilities that support request limits. */
     timeout?: number;
+    /**
+     * Optional asset snapshot to reuse instead of retrieving the asset from Mux.
+     * The snapshot must belong to `assetId`; callers should capture it immediately
+     * before starting the workflow.
+     */
+    assetSnapshot?: MuxAsset;
     /**
      * Optional credentials for workflow execution.
      * Use encryptForWorkflow when running in Workflow Dev Kit environments.

package/dist/workflows/index.d.ts

@@ -1,7 +1,7 @@
-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 StaticRenditionCleanupOutcome, a7 as SummarizationOptions, a8 as SummarizationPromptOverrides, a9 as SummarizationPromptSections, aa as SummaryAndTagsResult, ab as SummaryType, ac as ThumbnailModerationScore, ad as TranslationChunkingOptions, ae as TranslationOptions, af as TranslationPayload, ag as TranslationResult, ah as aggregateTokenUsage, ai as applyOverrideLists, aj as applyReplacements, ak as askQuestions, al as buildReplacementRegex, am as burnedInCaptionsSchema, an as censorVttContent, ao as chapterSchema, ap as chaptersSchema, aq as createReplacer, ar as editCaptions, as as extractTranscriptSegmentsForHotspots, at as generateChapters, au as generateEmbeddings, av as generateEngagementInsights, aw as generateVideoEmbeddings, ax as getModerationScores, ay as getSummaryAndTags, az as hasBurnedInCaptions, aA as mapShotsToHotspots, aB as normalizeTranslatedVtt, aC as profanityDetectionSchema, aD as questionAnswerSchema, aE as shouldSplitChunkTranslationError, aF as summarySchema, aG as transformCueText, aH as translateAudio, aI as translateCaptions, aJ as translationSchema, aK as validateNeverTranslateTerms, aL as verifyNeverTranslateTerms } from '../index-fU3hmKZu.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 SpeakerReplacement, a7 as StaticRenditionCleanupOutcome, a8 as SummarizationOptions, a9 as SummarizationPromptOverrides, aa as SummarizationPromptSections, ab as SummaryAndTagsResult, ac as SummaryType, ad as ThumbnailModerationScore, ae as TranslationChunkingOptions, af as TranslationOptions, ag as TranslationPayload, ah as TranslationResult, ai as aggregateTokenUsage, aj as applyOverrideLists, ak as applyReplacements, al as applySpeakerReplacements, am as askQuestions, an as buildReplacementRegex, ao as burnedInCaptionsSchema, ap as censorVttContent, aq as chapterSchema, ar as chaptersSchema, as as createReplacer, at as editCaptions, au as extractTranscriptSegmentsForHotspots, av as generateChapters, aw as generateEmbeddings, ax as generateEngagementInsights, ay as generateVideoEmbeddings, az as getModerationScores, aA as getSummaryAndTags, aB as hasBurnedInCaptions, aC as mapShotsToHotspots, aD as normalizeTranslatedVtt, aE as profanityDetectionSchema, aF as questionAnswerSchema, aG as shouldSplitChunkTranslationError, aH as summarySchema, aI as transformCueText, aJ as translateAudio, aK as translateCaptions, aL as translationSchema, aM as validateNeverTranslateTerms, aN as verifyNeverTranslateTerms } from '../index-BupL84pY.js';
 import 'zod';
 import '@ai-sdk/anthropic';
 import '@ai-sdk/google';
 import '@ai-sdk/openai';
-import '../shots-BMbViFbZ.js';
+import '../shots-CUI09Ohe.js';
 import '@mux/mux-node';

package/dist/workflows/index.js

@@ -1635,9 +1635,12 @@ function toPlaybackAsset(asset) {
   const { id: playbackId, policy } = getPlaybackId(asset);
   return { asset, playbackId, policy };
 }
-async function getPlaybackIdForAsset(assetId, credentials) {
+async function getPlaybackIdForAsset(assetId, credentials, assetSnapshot) {
   "use step";
-  const asset = await getMuxAsset(assetId, credentials);
+  const asset = assetSnapshot ?? await getMuxAsset(assetId, credentials);
+  if (asset.id !== assetId) {
+    throw new MuxAiError("Asset snapshot does not match the requested asset ID.", { type: "validation_error" });
+  }
   return toPlaybackAsset(asset);
 }
 async function getMuxAsset(assetId, credentials) {
@@ -3023,7 +3026,7 @@ async function askQuestionsInternal(assetId, questions, options, collectedUsage)
     model,
     provider
   });
-  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials);
+  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials, options?.assetSnapshot);
   const assetDurationSeconds = getAssetDurationSecondsFromAsset(assetData);
   const isAudioOnly = isAudioOnlyAsset(assetData);
   const effectiveScope = hasWorkflowScopeBoundaries(scope) ? scope : void 0;
@@ -3377,7 +3380,7 @@ async function hasBurnedInCaptionsInternal(assetId, options, collectedUsage) {
     model,
     provider
   });
-  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials);
+  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials, options.assetSnapshot);
   const assetDurationSeconds = getAssetDurationSecondsFromAsset(assetData);
   const storyboardScope = resolveRenderableVideoScope(
     scope,
@@ -3642,7 +3645,7 @@ async function generateChaptersInternal(assetId, options, collectedUsage) {
     model,
     provider
   });
-  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials);
+  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials, options.assetSnapshot);
   const assetDurationSeconds = getAssetDurationSecondsFromAsset(assetData);
   const effectiveScope = hasWorkflowScopeBoundaries(scope) ? scope : void 0;
   const resolvedScope = effectiveScope ? resolveWorkflowScope(effectiveScope, assetDurationSeconds) : void 0;
@@ -4204,6 +4207,27 @@ function applyReplacements(rawVtt, replacements) {
   });
   return { editedVtt, replacements: records };
 }
+function applySpeakerReplacements(rawVtt, speakerReplacements) {
+  const replacementsByLabel = new Map(
+    speakerReplacements.filter((replacement) => replacement.find.length > 0).map((replacement) => [replacement.find, replacement.replace])
+  );
+  if (replacementsByLabel.size === 0) {
+    return { editedVtt: rawVtt, replacements: [] };
+  }
+  const records = [];
+  const editedVtt = transformCueTextBlocks(rawVtt, (cueText, cueStartTime) => {
+    return cueText.replace(/^(\s*)\[([^\]\r\n]+)\](?=\s|$)/, (marker, leadingWhitespace, label) => {
+      const replacement = replacementsByLabel.get(label);
+      if (replacement === void 0)
+        return marker;
+      const before = `[${label}]`;
+      const after = `[${replacement}]`;
+      records.push({ cueStartTime, before, after });
+      return `${leadingWhitespace}${after}`;
+    });
+  });
+  return { editedVtt, replacements: records };
+}
 async function identifyProfanityWithAI({
   plainText,
   provider,
@@ -4306,6 +4330,7 @@ async function editCaptionsInternal(assetId, trackId, options, collectedUsage) {
     model,
     autoCensorProfanity: autoCensorOption,
     replacements: replacementsOption,
+    speakerReplacements: speakerReplacementsOption,
     deleteOriginalTrack,
     uploadToS3: uploadToS3Option,
     uploadToMux: uploadToMuxOption,
@@ -4318,8 +4343,9 @@ async function editCaptionsInternal(assetId, trackId, options, collectedUsage) {
   } = options;
   const hasAutoCensor = !!autoCensorOption;
   const hasReplacements = !!replacementsOption && replacementsOption.length > 0;
-  if (!hasAutoCensor && !hasReplacements) {
-    throw new MuxAiError("At least one of autoCensorProfanity or replacements must be provided.", { type: "validation_error" });
+  const hasSpeakerReplacements = !!speakerReplacementsOption && speakerReplacementsOption.length > 0;
+  if (!hasAutoCensor && !hasReplacements && !hasSpeakerReplacements) {
+    throw new MuxAiError("At least one of autoCensorProfanity, replacements, or speakerReplacements must be provided.", { type: "validation_error" });
   }
   if (autoCensorOption && !provider) {
     throw new MuxAiError("provider is required when using autoCensorProfanity.", { type: "validation_error" });
@@ -4327,6 +4353,16 @@ async function editCaptionsInternal(assetId, trackId, options, collectedUsage) {
   const deleteOriginal = deleteOriginalTrack !== false;
   const uploadToMux = uploadToMuxOption !== false;
   const uploadToS3 = uploadToS3Option || uploadToMux;
+  for (const replacement of speakerReplacementsOption ?? []) {
+    const hasInvalidFindCharacter = replacement.find.includes("[") || replacement.find.includes("]") || /[\r\n]/.test(replacement.find);
+    const hasInvalidReplaceCharacter = replacement.replace.includes("[") || replacement.replace.includes("]") || /[\r\n]/.test(replacement.replace);
+    if (!replacement.find || !replacement.replace || hasInvalidFindCharacter || hasInvalidReplaceCharacter) {
+      throw new MuxAiError(
+        "Speaker replacement labels must be non-empty and must not contain square brackets or newlines.",
+        { type: "validation_error" }
+      );
+    }
+  }
   const s3Endpoint = providedS3Endpoint ?? env_default.S3_ENDPOINT;
   const s3Region = providedS3Region ?? env_default.S3_REGION ?? "auto";
   const s3Bucket = providedS3Bucket ?? env_default.S3_BUCKET;
@@ -4338,7 +4374,7 @@ async function editCaptionsInternal(assetId, trackId, options, collectedUsage) {
       { type: "validation_error" }
     );
   }
-  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials);
+  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials, options.assetSnapshot);
   const assetDurationSeconds = getAssetDurationSecondsFromAsset(assetData);
   const signingContext = await resolveMuxSigningContext(credentials);
   if (policy === "signed" && !signingContext) {
@@ -4408,6 +4444,16 @@ async function editCaptionsInternal(assetId, trackId, options, collectedUsage) {
     totalReplacementCount += censorReplacements.length;
     autoCensorResult = { replacements: censorReplacements };
   }
+  let speakerReplacementsResult;
+  if (speakerReplacementsOption && speakerReplacementsOption.length > 0) {
+    const { editedVtt: afterSpeakerReplacements, replacements: speakerReplacements } = applySpeakerReplacements(
+      editedVtt,
+      speakerReplacementsOption
+    );
+    editedVtt = afterSpeakerReplacements;
+    totalReplacementCount += speakerReplacements.length;
+    speakerReplacementsResult = { replacements: speakerReplacements };
+  }
   let replacementsResult;
   if (replacementsOption && replacementsOption.length > 0) {
     const { editedVtt: afterReplacements, replacements: staticReplacements } = applyReplacements(editedVtt, replacementsOption);
@@ -4470,6 +4516,7 @@ async function editCaptionsInternal(assetId, trackId, options, collectedUsage) {
     totalReplacementCount,
     autoCensorProfanity: autoCensorResult,
     replacements: replacementsResult,
+    speakerReplacements: speakerReplacementsResult,
     uploadedTrackId,
     presignedUrl,
     usage: usageWithMetadata,
@@ -4774,7 +4821,7 @@ async function generateEmbeddingsInternal(assetId, options, collectedUsage) {
     scope
   } = options;
   const embeddingModel = resolveEmbeddingModelConfig({ ...options, provider, model });
-  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials);
+  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials, options.assetSnapshot);
   const assetDurationSeconds = getAssetDurationSecondsFromAsset(assetData);
   const effectiveScope = hasWorkflowScopeBoundaries(scope) ? scope : void 0;
   if (effectiveScope) {
@@ -5394,7 +5441,7 @@ async function generateEngagementInsightsInternal(assetId, options, collectedUsa
     model,
     provider
   });
-  const { asset, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials);
+  const { asset, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials, options.assetSnapshot);
   const assetDurationSeconds = getAssetDurationSecondsFromAsset(asset);
   if (!assetDurationSeconds) {
     throw new MuxAiError(`Asset ${assetId} has no valid duration.`, { type: "validation_error" });
@@ -6152,7 +6199,7 @@ async function getModerationScores(assetId, options = {}) {
     scope
   } = options;
   const credentials = providedCredentials;
-  const { asset, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials);
+  const { asset, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials, options.assetSnapshot);
   const videoTrackDurationSeconds = getVideoTrackDurationSecondsFromAsset(asset);
   const videoTrackFps = getVideoTrackMaxFrameRateFromAsset(asset);
   const assetDurationSeconds = getAssetDurationSecondsFromAsset(asset);
@@ -6766,7 +6813,7 @@ async function getSummaryAndTagsInternal(assetId, options, collectedUsage) {
     provider
   });
   const workflowCredentials = credentials;
-  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, workflowCredentials);
+  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, workflowCredentials, options?.assetSnapshot);
   const assetDurationSeconds = getAssetDurationSecondsFromAsset(assetData);
   const isAudioOnly = isAudioOnlyAsset(assetData);
   const effectiveScope = hasWorkflowScopeBoundaries(scope) ? scope : void 0;
@@ -7213,7 +7260,7 @@ async function translateAudio(assetId, toLanguageCode, options = {}) {
   if (uploadToS3 && (!s3Endpoint || !s3Bucket || !effectiveStorageAdapter && (!s3AccessKeyId || !s3SecretAccessKey))) {
     throw new MuxAiError("Storage configuration is required for uploading. Provide s3Endpoint and s3Bucket. If no storageAdapter is supplied, also provide s3AccessKeyId and s3SecretAccessKey in options or set S3_ENDPOINT, S3_BUCKET, S3_ACCESS_KEY_ID, and S3_SECRET_ACCESS_KEY environment variables.", { type: "validation_error" });
   }
-  const { asset: initialAsset, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials);
+  const { asset: initialAsset, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials, options.assetSnapshot);
   const assetDurationSeconds = getAssetDurationSecondsFromAsset(initialAsset);
   let currentAsset = initialAsset;
   let createdStaticRenditionId;
@@ -8083,7 +8130,7 @@ async function translateCaptionsInternal(assetId, trackId, toLanguageCode, optio
   if (uploadToS3 && (!s3Endpoint || !s3Bucket || !effectiveStorageAdapter && (!s3AccessKeyId || !s3SecretAccessKey))) {
     throw new MuxAiError("Storage configuration is required for uploading. Provide s3Endpoint and s3Bucket. If no storageAdapter is supplied, also provide s3AccessKeyId and s3SecretAccessKey in options or set S3_ENDPOINT, S3_BUCKET, S3_ACCESS_KEY_ID, and S3_SECRET_ACCESS_KEY environment variables.", { type: "validation_error" });
   }
-  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials);
+  const { asset: assetData, playbackId, policy } = await getPlaybackIdForAsset(assetId, credentials, options.assetSnapshot);
   const assetDurationSeconds = getAssetDurationSecondsFromAsset(assetData);
   const signingContext = await resolveMuxSigningContext(credentials);
   if (policy === "signed" && !signingContext) {
@@ -8236,6 +8283,7 @@ export {
   aggregateTokenUsage,
   applyOverrideLists,
   applyReplacements,
+  applySpeakerReplacements,
   askQuestions,
   buildReplacementRegex,
   burnedInCaptionsSchema,