@mux/ai 0.20.0 → 0.21.2

package/dist/{shots-BFYOfBgN.d.ts → shots-mGJfYGiO.d.ts}

@@ -120,6 +120,8 @@ interface WorkflowCredentials {
     anthropicApiKey?: string;
     /** Optional direct Google API key for per-request credential injection. */
     googleApiKey?: string;
+    /** Optional direct Google Vision API key for per-request credential injection. */
+    googleVisionApiKey?: string;
     /** Optional direct Hive API key for per-request credential injection. */
     hiveApiKey?: string;
     /** Optional direct ElevenLabs API key for per-request credential injection. */

package/dist/workflows/index.d.ts

@@ -1,7 +1,7 @@
-export { A as AskQuestionsOptions, a as AskQuestionsResult, b as AskQuestionsType, c as AudioTranslationOptions, d as AudioTranslationResult, e as AutoCensorProfanityOptions, B as BurnedInCaptionsAnalysis, f as BurnedInCaptionsOptions, g as BurnedInCaptionsPromptOverrides, h as BurnedInCaptionsPromptSections, j as BurnedInCaptionsResult, C as CaptionReplacement, k as CensorMode, l as Chapter, m as ChapterSystemPromptSections, n as ChaptersOptions, o as ChaptersPromptOverrides, p as ChaptersPromptSections, q as ChaptersResult, r as ChaptersType, D as DEFAULT_DESCRIPTION_LENGTH, s as DEFAULT_SUMMARY_KEYWORD_LIMIT, t as DEFAULT_TITLE_LENGTH, E as EditCaptionsOptions, u as EditCaptionsResult, v as EmbeddingsOptions, w as EmbeddingsResult, x as EngagementInsightsOptions, y as EngagementInsightsResult, H as HIVE_SEXUAL_CATEGORIES, z as HIVE_VIOLENCE_CATEGORIES, F as HiveModerationOutput, G as HiveModerationSource, M as ModerationOptions, I as ModerationProvider, J as ModerationResult, K as MomentInsight, O as OverallInsight, P as ProfanityDetectionPayload, Q as Question, L as QuestionAnswer, N as QuestionAnswerType, R as ReplacementRecord, S as SummarizationOptions, T as SummarizationPromptOverrides, U as SummarizationPromptSections, V as SummaryAndTagsResult, W as SummaryType, X as ThumbnailModerationScore, Y as TranslationChunkingOptions, Z as TranslationOptions, _ as TranslationPayload, $ as TranslationResult, a0 as aggregateTokenUsage, a1 as applyOverrideLists, a2 as applyReplacements, a3 as askQuestions, a4 as buildReplacementRegex, a5 as burnedInCaptionsSchema, a6 as censorVttContent, a7 as chapterSchema, a8 as chaptersSchema, a9 as createReplacer, aa as editCaptions, ab as extractTranscriptSegmentsForHotspots, ac as generateChapters, ad as generateEmbeddings, ae as generateEngagementInsights, af as generateVideoEmbeddings, ag as getModerationScores, ah as getSummaryAndTags, ai as hasBurnedInCaptions, aj as mapShotsToHotspots, ak as normalizeTranslatedVtt, al as profanityDetectionSchema, am as questionAnswerSchema, an as shouldSplitChunkTranslationError, ao as summarySchema, ap as transformCueText, aq as translateAudio, ar as translateCaptions, as as translationSchema } from '../index-DXMIJjUa.js';
+export { A as AskQuestionsOptions, a as AskQuestionsResult, b as AskQuestionsType, c as AudioTranslationOptions, d as AudioTranslationResult, e as AutoCensorProfanityOptions, B as BurnedInCaptionsAnalysis, f as BurnedInCaptionsOptions, g as BurnedInCaptionsPromptOverrides, h as BurnedInCaptionsPromptSections, j as BurnedInCaptionsResult, C as CaptionReplacement, k as CensorMode, l as Chapter, m as ChapterSystemPromptSections, n as ChaptersOptions, o as ChaptersPromptOverrides, p as ChaptersPromptSections, q as ChaptersResult, r as ChaptersType, D as DEFAULT_DESCRIPTION_LENGTH, s as DEFAULT_SUMMARY_KEYWORD_LIMIT, t as DEFAULT_TITLE_LENGTH, E as EditCaptionsOptions, u as EditCaptionsResult, v as EmbeddingsOptions, w as EmbeddingsResult, x as EngagementInsightsOptions, y as EngagementInsightsResult, G as GOOGLE_VISION_LIKELIHOOD_TO_SCORE, H as HIVE_SEXUAL_CATEGORIES, z as HIVE_VIOLENCE_CATEGORIES, F as HiveModerationOutput, I as HiveModerationSource, M as ModerationOptions, J as ModerationProvider, K as ModerationResult, L as MomentInsight, O as OverallInsight, P as ProfanityDetectionPayload, Q as Question, N as QuestionAnswer, R as QuestionAnswerType, S as ReplacementRecord, T as SummarizationOptions, U as SummarizationPromptOverrides, V as SummarizationPromptSections, W as SummaryAndTagsResult, X as SummaryType, Y as ThumbnailModerationScore, Z as TranslationChunkingOptions, _ as TranslationOptions, $ as TranslationPayload, a0 as TranslationResult, a1 as aggregateTokenUsage, a2 as applyOverrideLists, a3 as applyReplacements, a4 as askQuestions, a5 as buildReplacementRegex, a6 as burnedInCaptionsSchema, a7 as censorVttContent, a8 as chapterSchema, a9 as chaptersSchema, aa as createReplacer, ab as editCaptions, ac as extractTranscriptSegmentsForHotspots, ad as generateChapters, ae as generateEmbeddings, af as generateEngagementInsights, ag as generateVideoEmbeddings, ah as getModerationScores, ai as getSummaryAndTags, aj as hasBurnedInCaptions, ak as mapShotsToHotspots, al as normalizeTranslatedVtt, am as profanityDetectionSchema, an as questionAnswerSchema, ao as shouldSplitChunkTranslationError, ap as summarySchema, aq as transformCueText, ar as translateAudio, as as translateCaptions, at as translationSchema } from '../index-DwNtsg-u.js';
 import 'zod';
 import '@ai-sdk/anthropic';
 import '@ai-sdk/google';
 import '@ai-sdk/openai';
-import '../shots-BFYOfBgN.js';
+import '../shots-mGJfYGiO.js';
 import '@mux/mux-node';

package/dist/workflows/index.js

@@ -186,6 +186,7 @@ var EnvSchema = z.object({
   GOOGLE_GENERATIVE_AI_API_KEY: optionalString("Google Generative AI API key for Gemini-backed workflows.", "Google Generative AI API key"),
   ELEVENLABS_API_KEY: optionalString("ElevenLabs API key for audio translation.", "ElevenLabs API key"),
   HIVE_API_KEY: optionalString("Hive Visual Moderation API key.", "Hive API key"),
+  GOOGLE_VISION_API_KEY: optionalString("Google Vision API key for SafeSearch moderation.", "Google Vision API key"),
   // S3-Compatible Storage (required for translation & audio dubbing)
   S3_ENDPOINT: optionalString("S3-compatible endpoint for uploads.", "S3 endpoint"),
   S3_REGION: optionalString("S3 region (defaults to 'auto' when omitted)."),
@@ -809,23 +810,26 @@ function resolveProviderApiKeyFromCredentials(provider, resolved) {
   const openaiApiKey = readString(record, "openaiApiKey");
   const anthropicApiKey = readString(record, "anthropicApiKey");
   const googleApiKey = readString(record, "googleApiKey");
+  const googleVisionApiKey = readString(record, "googleVisionApiKey");
   const hiveApiKey = readString(record, "hiveApiKey");
   const elevenLabsApiKey = readString(record, "elevenLabsApiKey");
   const apiKeyMap = {
-    openai: openaiApiKey ?? env_default.OPENAI_API_KEY,
-    anthropic: anthropicApiKey ?? env_default.ANTHROPIC_API_KEY,
-    google: googleApiKey ?? env_default.GOOGLE_GENERATIVE_AI_API_KEY,
-    hive: hiveApiKey ?? env_default.HIVE_API_KEY,
-    elevenlabs: elevenLabsApiKey ?? env_default.ELEVENLABS_API_KEY
+    "openai": openaiApiKey ?? env_default.OPENAI_API_KEY,
+    "anthropic": anthropicApiKey ?? env_default.ANTHROPIC_API_KEY,
+    "google": googleApiKey ?? env_default.GOOGLE_GENERATIVE_AI_API_KEY,
+    "google-vision-api": googleVisionApiKey ?? env_default.GOOGLE_VISION_API_KEY,
+    "hive": hiveApiKey ?? env_default.HIVE_API_KEY,
+    "elevenlabs": elevenLabsApiKey ?? env_default.ELEVENLABS_API_KEY
   };
   const apiKey = apiKeyMap[provider];
   if (!apiKey) {
     const envVarNames = {
-      openai: "OPENAI_API_KEY",
-      anthropic: "ANTHROPIC_API_KEY",
-      google: "GOOGLE_GENERATIVE_AI_API_KEY",
-      hive: "HIVE_API_KEY",
-      elevenlabs: "ELEVENLABS_API_KEY"
+      "openai": "OPENAI_API_KEY",
+      "anthropic": "ANTHROPIC_API_KEY",
+      "google": "GOOGLE_GENERATIVE_AI_API_KEY",
+      "google-vision-api": "GOOGLE_VISION_API_KEY",
+      "hive": "HIVE_API_KEY",
+      "elevenlabs": "ELEVENLABS_API_KEY"
     };
     throw new Error(
       `${provider} API key is required. Provide ${provider} credentials via workflow credentials or set ${envVarNames[provider]} environment variable.`
@@ -861,17 +865,26 @@ var DEFAULT_EMBEDDING_MODELS = {
 var LANGUAGE_MODELS = {
   openai: ["gpt-5.1", "gpt-5-mini"],
   anthropic: ["claude-sonnet-4-5"],
-  google: ["gemini-3-flash-preview", "gemini-3.1-flash-lite-preview", "gemini-2.5-flash"]
+  google: ["gemini-3-flash-preview", "gemini-3.1-flash-lite", "gemini-3.1-flash-lite-preview", "gemini-2.5-flash"]
 };
 var LANGUAGE_MODEL_DEPRECATIONS = [
   {
     provider: "google",
     modelId: "gemini-2.5-flash",
-    replacementModelId: "gemini-3.1-flash-lite-preview",
+    replacementModelId: "gemini-3.1-flash-lite",
     phase: "warn",
     deprecatedOn: "2026-03-03",
-    sunsetOn: "2026-06-30",
-    reason: "Gemini 3.1 Flash-Lite Preview offers better quality/latency/cost balance in current evals."
+    sunsetOn: "2026-06-17",
+    reason: "Gemini 3.1 Flash-Lite offers better quality/latency/cost balance in current evals."
+  },
+  {
+    provider: "google",
+    modelId: "gemini-3.1-flash-lite-preview",
+    replacementModelId: "gemini-3.1-flash-lite",
+    phase: "warn",
+    deprecatedOn: "2026-05-12",
+    sunsetOn: "2026-05-25",
+    reason: "Preview model graduated to the stable Gemini 3.1 Flash-Lite release."
   }
 ];
 var warnedDeprecatedLanguageModels = /* @__PURE__ */ new Set();
@@ -4859,6 +4872,15 @@ var OPENAI_MODERATION_BASE_DELAY_MS = 750;
 var OPENAI_MODERATION_MAX_DELAY_MS = 3e3;
 var MIN_SAMPLE_COVERAGE_FOR_CONFIDENT_THRESHOLDING = 0.5;
 var MIN_SUCCESSFUL_THUMBNAILS_FOR_CONFIDENT_THRESHOLDING = 3;
+var GOOGLE_VISION_ENDPOINT = "https://vision.googleapis.com/v1/images:annotate";
+var GOOGLE_VISION_LIKELIHOOD_TO_SCORE = {
+  UNKNOWN: 0,
+  VERY_UNLIKELY: 0.2,
+  UNLIKELY: 0.4,
+  POSSIBLE: 0.6,
+  LIKELY: 0.8,
+  VERY_LIKELY: 1
+};
 var HIVE_ENDPOINT = "https://api.thehive.ai/api/v2/task/sync";
 var HIVE_SEXUAL_CATEGORIES = [
   "general_nsfw",
@@ -5187,6 +5209,98 @@ async function requestHiveModeration(images, maxConcurrent = 5, submissionMode =
   }));
   return await processConcurrently(targets, moderateImageWithHive, maxConcurrent);
 }
+async function moderateImageWithGoogleVision(entry) {
+  "use step";
+  try {
+    const apiKey = await getApiKeyFromEnv("google-vision-api", entry.credentials);
+    const imageField = entry.isBase64 ? { content: entry.image } : { source: { imageUri: entry.image } };
+    const controller = new AbortController();
+    const timeout = setTimeout(() => controller.abort(), 15e3);
+    let res;
+    try {
+      res = await fetch(GOOGLE_VISION_ENDPOINT, {
+        method: "POST",
+        headers: {
+          "Accept": "application/json",
+          "Content-Type": "application/json",
+          "X-goog-api-key": apiKey
+        },
+        body: JSON.stringify({
+          requests: [
+            {
+              image: imageField,
+              features: [{ type: "SAFE_SEARCH_DETECTION" }]
+            }
+          ]
+        }),
+        signal: controller.signal
+      });
+    } catch (err) {
+      if (err?.name === "AbortError") {
+        throw new Error("Google Vision request timed out after 15s");
+      }
+      throw err;
+    } finally {
+      clearTimeout(timeout);
+    }
+    const json = await res.json().catch(() => void 0);
+    if (!res.ok) {
+      throw new Error(
+        `Google Vision moderation error: ${res.status} ${res.statusText} - ${JSON.stringify(json)}`
+      );
+    }
+    const perImage = json?.responses?.[0];
+    if (perImage?.error) {
+      throw new Error(
+        `Google Vision per-image error: ${perImage.error.code} ${perImage.error.message || "Unknown error"}`
+      );
+    }
+    const annotation = perImage?.safeSearchAnnotation;
+    if (!annotation) {
+      throw new Error(`Google Vision response missing safeSearchAnnotation: ${JSON.stringify(json)}`);
+    }
+    const adultLikelihood = typeof annotation.adult === "string" ? annotation.adult : "UNKNOWN";
+    const violenceLikelihood = typeof annotation.violence === "string" ? annotation.violence : "UNKNOWN";
+    const sexual = GOOGLE_VISION_LIKELIHOOD_TO_SCORE[adultLikelihood] ?? 0;
+    const violence = GOOGLE_VISION_LIKELIHOOD_TO_SCORE[violenceLikelihood] ?? 0;
+    return {
+      url: entry.url,
+      time: entry.time,
+      sexual,
+      violence,
+      error: false
+    };
+  } catch (error) {
+    return {
+      url: entry.url,
+      time: entry.time,
+      sexual: 0,
+      violence: 0,
+      error: true,
+      errorMessage: error instanceof Error ? error.message : String(error)
+    };
+  }
+}
+async function requestGoogleVisionModeration(images, maxConcurrent = 5, submissionMode = "url", downloadOptions, credentials) {
+  "use step";
+  const imageUrls = images.map((img) => img.url);
+  const timeByUrl = new Map(images.map((img) => [img.url, img.time]));
+  const targets = submissionMode === "base64" ? (await downloadImagesAsBase64(imageUrls, downloadOptions, maxConcurrent)).map((img) => ({
+    url: img.url,
+    time: timeByUrl.get(img.url),
+    // Vision REST wants raw base64; downloadImagesAsBase64 returns a data URI.
+    image: img.base64Data.startsWith("data:") ? img.base64Data.split(",")[1] : img.base64Data,
+    isBase64: true,
+    credentials
+  })) : images.map((img) => ({
+    url: img.url,
+    time: img.time,
+    image: img.url,
+    isBase64: false,
+    credentials
+  }));
+  return processConcurrently(targets, moderateImageWithGoogleVision, maxConcurrent);
+}
 async function getThumbnailUrlsFromTimestamps(playbackId, timestampsMs, options) {
   "use step";
   const { width, shouldSign, credentials } = options;
@@ -5250,8 +5364,11 @@ async function getModerationScores(assetId, options = {}) {
       );
     } else if (provider === "hive") {
       throw new Error("Hive does not support transcript moderation in this workflow. Use provider: 'openai' for audio-only assets.");
+    } else if (provider === "google-vision-api") {
+      throw new Error("google-vision-api is image-only and does not support transcript moderation. Use provider: 'openai' for audio-only assets.");
     } else {
-      throw new Error(`Unsupported moderation provider: ${provider}`);
+      const exhaustiveCheck = provider;
+      throw new Error(`Unsupported moderation provider: ${exhaustiveCheck}`);
     }
   } else {
     const estimatedIntervalCount = duration <= 50 ? 5 : Math.ceil(duration / thumbnailInterval);
@@ -5294,8 +5411,17 @@ async function getModerationScores(assetId, options = {}) {
         imageDownloadOptions,
         credentials
       );
+    } else if (provider === "google-vision-api") {
+      thumbnailScores = await requestGoogleVisionModeration(
+        thumbnailUrls,
+        maxConcurrent,
+        imageSubmissionMode,
+        imageDownloadOptions,
+        credentials
+      );
     } else {
-      throw new Error(`Unsupported moderation provider: ${provider}`);
+      const exhaustiveCheck = provider;
+      throw new Error(`Unsupported moderation provider: ${exhaustiveCheck}`);
     }
   }
   const failed = thumbnailScores.filter((s) => s.error);
@@ -6999,6 +7125,7 @@ export {
   DEFAULT_DESCRIPTION_LENGTH,
   DEFAULT_SUMMARY_KEYWORD_LIMIT,
   DEFAULT_TITLE_LENGTH,
+  GOOGLE_VISION_LIKELIHOOD_TO_SCORE,
   HIVE_SEXUAL_CATEGORIES,
   HIVE_VIOLENCE_CATEGORIES,
   aggregateTokenUsage,