@mux/ai 0.5.0 → 0.5.2

package/dist/{index-Dmo55n-5.d.ts → index-2oUwgWsE.d.ts}

@@ -1,4 +1,4 @@
-import { b as WorkflowCredentialsInput, A as AssetTextTrack, c as MuxAsset, h as TextChunk, C as ChunkingStrategy } from './types-KcVfWtUl.js';
+import { b as WorkflowCredentialsInput, A as AssetTextTrack, c as MuxAsset, h as TextChunk, C as ChunkingStrategy } from './types-BhVuLeSp.js';
 
 declare const DEFAULT_STORYBOARD_WIDTH = 640;
 /**
@@ -110,6 +110,8 @@ interface ThumbnailOptions {
     width?: number;
     /** Flag for whether or not to use signed playback IDs (default: false) */
     shouldSign?: boolean;
+    /** Maximum number of thumbnails to generate. When set, samples are evenly distributed with first and last frames pinned. */
+    maxSamples?: number;
     /** Workflow credentials for signing (optional). */
     credentials?: WorkflowCredentialsInput;
 }

package/dist/{index-BdpEVCbj.d.ts → index-DgCfxP4T.d.ts}

@@ -2,7 +2,7 @@ import { z } from 'zod';
 import { createAnthropic } from '@ai-sdk/anthropic';
 import { createGoogleGenerativeAI } from '@ai-sdk/google';
 import { createOpenAI } from '@ai-sdk/openai';
-import { k as TokenUsage, M as MuxAIOptions, I as ImageSubmissionMode, C as ChunkingStrategy, j as VideoEmbeddingsResult, T as ToneType } from './types-KcVfWtUl.js';
+import { k as TokenUsage, M as MuxAIOptions, I as ImageSubmissionMode, C as ChunkingStrategy, j as VideoEmbeddingsResult, T as ToneType } from './types-BhVuLeSp.js';
 
 interface ImageDownloadOptions {
     /** Request timeout in milliseconds (default: 10000) */
@@ -124,11 +124,11 @@ declare const chapterSchema: z.ZodObject<{
     startTime: z.ZodNumber;
     title: z.ZodString;
 }, "strip", z.ZodTypeAny, {
-    startTime: number;
     title: string;
-}, {
     startTime: number;
+}, {
     title: string;
+    startTime: number;
 }>;
 type Chapter = z.infer<typeof chapterSchema>;
 declare const chaptersSchema: z.ZodObject<{
@@ -136,21 +136,21 @@ declare const chaptersSchema: z.ZodObject<{
         startTime: z.ZodNumber;
         title: z.ZodString;
     }, "strip", z.ZodTypeAny, {
-        startTime: number;
         title: string;
-    }, {
         startTime: number;
+    }, {
         title: string;
+        startTime: number;
     }>, "many">;
 }, "strip", z.ZodTypeAny, {
     chapters: {
-        startTime: number;
         title: string;
+        startTime: number;
     }[];
 }, {
     chapters: {
-        startTime: number;
         title: string;
+        startTime: number;
     }[];
 }>;
 type ChaptersType = z.infer<typeof chaptersSchema>;
@@ -292,6 +292,8 @@ interface ModerationOptions extends MuxAIOptions {
     thumbnailInterval?: number;
     /** Width of storyboard thumbnails in pixels (defaults to 640). */
     thumbnailWidth?: number;
+    /** Maximum number of thumbnails to sample (defaults to unlimited). When set, samples are evenly distributed with first and last frames pinned. */
+    maxSamples?: number;
     /** Max concurrent moderation requests (defaults to 5). */
     maxConcurrent?: number;
     /** Transport used for thumbnails (defaults to 'url'). */
@@ -318,12 +320,12 @@ declare const summarySchema: z.ZodObject<{
     description: z.ZodString;
 }, "strip", z.ZodTypeAny, {
     title: string;
-    keywords: string[];
     description: string;
+    keywords: string[];
 }, {
     title: string;
-    keywords: string[];
     description: string;
+    keywords: string[];
 }>;
 type SummaryType = z.infer<typeof summarySchema>;
 /** Structured return payload for `getSummaryAndTags`. */

package/dist/index.d.ts

@@ -1,14 +1,14 @@
-import { W as WorkflowCredentials } from './types-KcVfWtUl.js';
-export { A as AssetTextTrack, i as ChunkEmbedding, C as ChunkingStrategy, E as Encrypted, a as EncryptedPayload, I as ImageSubmissionMode, M as MuxAIOptions, c as MuxAsset, f as PlaybackAsset, P as PlaybackPolicy, h as TextChunk, g as TokenChunkingConfig, k as TokenUsage, T as ToneType, V as VTTChunkingConfig, j as VideoEmbeddingsResult, b as WorkflowCredentialsInput, d as decryptFromWorkflow, e as encryptForWorkflow } from './types-KcVfWtUl.js';
-export { i as primitives } from './index-Dmo55n-5.js';
-export { i as workflows } from './index-BdpEVCbj.js';
+import { W as WorkflowCredentials } from './types-BhVuLeSp.js';
+export { A as AssetTextTrack, i as ChunkEmbedding, C as ChunkingStrategy, E as Encrypted, a as EncryptedPayload, I as ImageSubmissionMode, M as MuxAIOptions, c as MuxAsset, f as PlaybackAsset, P as PlaybackPolicy, h as TextChunk, g as TokenChunkingConfig, k as TokenUsage, T as ToneType, V as VTTChunkingConfig, j as VideoEmbeddingsResult, b as WorkflowCredentialsInput, d as decryptFromWorkflow, e as encryptForWorkflow } from './types-BhVuLeSp.js';
+export { i as primitives } from './index-2oUwgWsE.js';
+export { i as workflows } from './index-DgCfxP4T.js';
 import '@mux/mux-node';
 import 'zod';
 import '@ai-sdk/anthropic';
 import '@ai-sdk/google';
 import '@ai-sdk/openai';
 
-var version = "0.5.0";
+var version = "0.5.2";
 
 /**
  * A function that returns workflow credentials, either synchronously or asynchronously.

package/dist/index.js

@@ -28018,7 +28018,7 @@ var init_dist10 = __esm({
 });
 
 // package.json
-var version = "0.5.0";
+var version = "0.5.2";
 
 // src/env.ts
 import { z } from "zod";
@@ -28534,8 +28534,8 @@ function chunkText(text, strategy) {
 // src/primitives/thumbnails.ts
 async function getThumbnailUrls(playbackId, duration, options = {}) {
   "use step";
-  const { interval = 10, width = 640, shouldSign = false, credentials } = options;
-  const timestamps = [];
+  const { interval = 10, width = 640, shouldSign = false, maxSamples, credentials } = options;
+  let timestamps = [];
   if (duration <= 50) {
     const spacing = duration / 6;
     for (let i = 1; i <= 5; i++) {
@@ -28546,6 +28546,18 @@ async function getThumbnailUrls(playbackId, duration, options = {}) {
       timestamps.push(time);
     }
   }
+  if (maxSamples !== void 0 && timestamps.length > maxSamples) {
+    const newTimestamps = [];
+    newTimestamps.push(0);
+    if (maxSamples >= 2) {
+      const spacing = duration / (maxSamples - 1);
+      for (let i = 1; i < maxSamples - 1; i++) {
+        newTimestamps.push(spacing * i);
+      }
+      newTimestamps.push(duration);
+    }
+    timestamps = newTimestamps;
+  }
   const baseUrl = `https://image.mux.com/${playbackId}/thumbnail.png`;
   const urlPromises = timestamps.map(async (time) => {
     if (shouldSign) {
@@ -28753,7 +28765,7 @@ __export(workflows_exports, {
 });
 
 // src/workflows/burned-in-captions.ts
-import { generateObject } from "ai";
+import { generateText, Output } from "ai";
 import dedent from "dedent";
 import { z as z21 } from "zod";
 
@@ -29049,7 +29061,7 @@ function createToneSection(instruction) {
 // src/workflows/burned-in-captions.ts
 var burnedInCaptionsSchema = z21.object({
   hasBurnedInCaptions: z21.boolean(),
-  confidence: z21.number().min(0).max(1),
+  confidence: z21.number(),
   detectedLanguage: z21.string().nullable()
 });
 var SYSTEM_PROMPT = dedent`
@@ -29150,9 +29162,9 @@ async function analyzeStoryboard({
 }) {
   "use step";
   const model = await createLanguageModelFromConfig(provider, modelId, credentials);
-  const response = await generateObject({
+  const response = await generateText({
     model,
-    schema: burnedInCaptionsSchema,
+    output: Output.object({ schema: burnedInCaptionsSchema }),
     experimental_telemetry: { isEnabled: true },
     messages: [
       {
@@ -29169,7 +29181,10 @@ async function analyzeStoryboard({
     ]
   });
   return {
-    result: response.object,
+    result: {
+      ...response.output,
+      confidence: Math.min(1, Math.max(0, response.output.confidence))
+    },
     usage: {
       inputTokens: response.usage.inputTokens,
       outputTokens: response.usage.outputTokens,
@@ -29233,7 +29248,7 @@ async function hasBurnedInCaptions(assetId, options = {}) {
 }
 
 // src/workflows/chapters.ts
-import { generateObject as generateObject2 } from "ai";
+import { generateText as generateText2, Output as Output2 } from "ai";
 import dedent2 from "dedent";
 import { z as z22 } from "zod";
 
@@ -29295,9 +29310,9 @@ async function generateChaptersWithAI({
   "use step";
   const model = await createLanguageModelFromConfig(provider, modelId, credentials);
   const response = await withRetry(
-    () => generateObject2({
+    () => generateText2({
       model,
-      schema: chaptersSchema,
+      output: Output2.object({ schema: chaptersSchema }),
       messages: [
         {
           role: "system",
@@ -29311,7 +29326,7 @@ async function generateChaptersWithAI({
     })
   );
   return {
-    chapters: response.object,
+    chapters: response.output,
     usage: {
       inputTokens: response.usage.inputTokens,
       outputTokens: response.usage.outputTokens,
@@ -29910,6 +29925,7 @@ async function getModerationScores(assetId, options = {}) {
     thresholds = DEFAULT_THRESHOLDS,
     thumbnailInterval = 10,
     thumbnailWidth = 640,
+    maxSamples,
     maxConcurrent = 5,
     imageSubmissionMode = "url",
     imageDownloadOptions,
@@ -29961,6 +29977,7 @@ async function getModerationScores(assetId, options = {}) {
       interval: thumbnailInterval,
       width: thumbnailWidth,
       shouldSign: policy === "signed",
+      maxSamples,
       credentials
     });
     if (provider === "openai") {
@@ -30002,7 +30019,7 @@ async function getModerationScores(assetId, options = {}) {
 }
 
 // src/workflows/summarization.ts
-import { generateObject as generateObject3 } from "ai";
+import { generateText as generateText3, Output as Output3 } from "ai";
 import dedent3 from "dedent";
 import { z as z23 } from "zod";
 var SUMMARY_KEYWORD_LIMIT = 10;
@@ -30232,9 +30249,9 @@ function buildUserPrompt3({
 async function analyzeStoryboard2(imageDataUrl, provider, modelId, userPrompt, systemPrompt, credentials) {
   "use step";
   const model = await createLanguageModelFromConfig(provider, modelId, credentials);
-  const response = await generateObject3({
+  const response = await generateText3({
     model,
-    schema: summarySchema,
+    output: Output3.object({ schema: summarySchema }),
     messages: [
       {
         role: "system",
@@ -30250,7 +30267,7 @@ async function analyzeStoryboard2(imageDataUrl, provider, modelId, userPrompt, s
     ]
   });
   return {
-    result: response.object,
+    result: response.output,
     usage: {
       inputTokens: response.usage.inputTokens,
       outputTokens: response.usage.outputTokens,
@@ -30263,9 +30280,9 @@ async function analyzeStoryboard2(imageDataUrl, provider, modelId, userPrompt, s
 async function analyzeAudioOnly(provider, modelId, userPrompt, systemPrompt, credentials) {
   "use step";
   const model = await createLanguageModelFromConfig(provider, modelId, credentials);
-  const response = await generateObject3({
+  const response = await generateText3({
     model,
-    schema: summarySchema,
+    output: Output3.object({ schema: summarySchema }),
     messages: [
       {
         role: "system",
@@ -30278,7 +30295,7 @@ async function analyzeAudioOnly(provider, modelId, userPrompt, systemPrompt, cre
     ]
   });
   return {
-    result: response.object,
+    result: response.output,
     usage: {
       inputTokens: response.usage.inputTokens,
       outputTokens: response.usage.outputTokens,
@@ -30321,7 +30338,6 @@ async function getSummaryAndTags(assetId, options) {
     cleanTranscript = true,
     imageSubmissionMode = "url",
     imageDownloadOptions,
-    abortSignal: _abortSignal,
     promptOverrides,
     credentials
   } = options ?? {};
@@ -30982,7 +30998,7 @@ async function translateAudio(assetId, toLanguageCode, options = {}) {
 
 // src/workflows/translate-captions.ts
 import Mux4 from "@mux/mux-node";
-import { generateObject as generateObject4 } from "ai";
+import { generateText as generateText4, Output as Output4 } from "ai";
 import { z as z24 } from "zod";
 var translationSchema = z24.object({
   translation: z24.string()
@@ -31001,15 +31017,13 @@ async function translateVttWithAI({
   toLanguageCode,
   provider,
   modelId,
-  abortSignal,
   credentials
 }) {
   "use step";
-  const languageModel = await createLanguageModelFromConfig(provider, modelId, credentials);
-  const response = await generateObject4({
-    model: languageModel,
-    schema: translationSchema,
-    abortSignal,
+  const model = await createLanguageModelFromConfig(provider, modelId, credentials);
+  const response = await generateText4({
+    model,
+    output: Output4.object({ schema: translationSchema }),
     messages: [
       {
         role: "user",
@@ -31020,7 +31034,7 @@ ${vttContent}`
     ]
   });
   return {
-    translatedVtt: response.object.translation,
+    translatedVtt: response.output.translation,
     usage: {
       inputTokens: response.usage.inputTokens,
       outputTokens: response.usage.outputTokens,
@@ -31167,7 +31181,6 @@ async function translateCaptions(assetId, fromLanguageCode, toLanguageCode, opti
       toLanguageCode,
       provider: modelConfig.provider,
       modelId: modelConfig.modelId,
-      abortSignal: options.abortSignal,
       credentials
     });
     translatedVtt = result.translatedVtt;