@mux/ai 0.7.5 → 0.8.0

package/dist/workflows/index.js

@@ -138,6 +138,10 @@ var EnvSchema = z.object({
   ),
   MUX_SIGNING_KEY: optionalString("Mux signing key ID for signed playback URLs.", "Used to sign playback URLs"),
   MUX_PRIVATE_KEY: optionalString("Mux signing private key for signed playback URLs.", "Used to sign playback URLs"),
+  MUX_IMAGE_URL_OVERRIDE: optionalString(
+    "Override for Mux image base URL (defaults to https://image.mux.com).",
+    "Mux image URL override"
+  ),
   // Test-only helpers (used by this repo's integration tests)
   MUX_TEST_ASSET_ID: optionalString("Mux asset ID used by integration tests.", "Mux test asset id"),
   MUX_TEST_ASSET_ID_CHAPTERS: optionalString("Mux asset ID used by integration tests for chapters.", "Mux test asset id for chapters"),
@@ -835,6 +839,44 @@ async function withRetry(fn, {
   throw lastError || new Error("Retry failed with unknown error");
 }
 
+// src/lib/mux-image-url.ts
+var DEFAULT_MUX_IMAGE_ORIGIN = "https://image.mux.com";
+function normalizeMuxImageOrigin(value) {
+  const trimmed = value.trim();
+  const candidate = trimmed.includes("://") ? trimmed : `https://${trimmed}`;
+  let parsed;
+  try {
+    parsed = new URL(candidate);
+  } catch {
+    throw new Error(
+      `Invalid MUX_IMAGE_URL_OVERRIDE. Provide a hostname like "image.example.mux.com" (or a URL origin such as "https://image.example.mux.com").`
+    );
+  }
+  if (parsed.username || parsed.password || parsed.search || parsed.hash || parsed.pathname && parsed.pathname !== "/") {
+    throw new Error(
+      "Invalid MUX_IMAGE_URL_OVERRIDE. Only a hostname/origin is allowed (no credentials, query params, hash fragments, or path)."
+    );
+  }
+  return parsed.origin;
+}
+function getMuxImageOrigin() {
+  const override = env_default.MUX_IMAGE_URL_OVERRIDE;
+  if (!override) {
+    return DEFAULT_MUX_IMAGE_ORIGIN;
+  }
+  return normalizeMuxImageOrigin(override);
+}
+function getMuxImageBaseUrl(playbackId, assetType) {
+  const origin = getMuxImageOrigin();
+  return `${origin}/${playbackId}/${assetType}.png`;
+}
+function getMuxStoryboardBaseUrl(playbackId) {
+  return getMuxImageBaseUrl(playbackId, "storyboard");
+}
+function getMuxThumbnailBaseUrl(playbackId) {
+  return getMuxImageBaseUrl(playbackId, "thumbnail");
+}
+
 // src/lib/url-signing.ts
 async function createSigningClient(context) {
   const { default: MuxClient } = await import("@mux/mux-node");
@@ -876,7 +918,7 @@ async function signUrl(url, playbackId, type = "video", params, credentials) {
 var DEFAULT_STORYBOARD_WIDTH = 640;
 async function getStoryboardUrl(playbackId, width = DEFAULT_STORYBOARD_WIDTH, shouldSign = false, credentials) {
   "use step";
-  const baseUrl = `https://image.mux.com/${playbackId}/storyboard.png`;
+  const baseUrl = getMuxStoryboardBaseUrl(playbackId);
   if (shouldSign) {
     return signUrl(baseUrl, playbackId, "storyboard", { width }, credentials);
   }
@@ -1128,16 +1170,6 @@ var SYSTEM_PROMPT = dedent`
     - GOOD: "A person runs through a park"
     - Be specific and evidence-based
   </language_guidelines>`;
-function buildSystemPrompt(allowedAnswers) {
-  const answerList = allowedAnswers.map((answer) => `"${answer}"`).join(", ");
-  return `${SYSTEM_PROMPT}
-
-${dedent`
-    <response_options>
-      Allowed answers: ${answerList}
-    </response_options>
-  `}`;
-}
 var askQuestionsPromptBuilder = createPromptBuilder({
   template: {
     questions: {
@@ -1147,21 +1179,30 @@ var askQuestionsPromptBuilder = createPromptBuilder({
   },
   sectionOrder: ["questions"]
 });
-function buildUserPrompt(questions, transcriptText, isCleanTranscript = true) {
+function buildUserPrompt(questions, allowedAnswers, transcriptText, isCleanTranscript = true) {
   const questionsList = questions.map((q, idx) => `${idx + 1}. ${q.question}`).join("\n");
   const questionsContent = dedent`
     Please answer the following yes/no questions about this video:
 
     ${questionsList}`;
+  const answerList = allowedAnswers.map((answer) => `"${answer}"`).join(", ");
+  const responseOptions = dedent`
+    <response_options>
+      Allowed answers: ${answerList}
+    </response_options>`;
+  const questionsSection = askQuestionsPromptBuilder.build({ questions: questionsContent });
   if (!transcriptText) {
-    return askQuestionsPromptBuilder.build({ questions: questionsContent });
+    return `${questionsSection}
+
+${responseOptions}`;
   }
   const format = isCleanTranscript ? "plain text" : "WebVTT";
-  const transcriptSection = createTranscriptSection(transcriptText, format);
-  return askQuestionsPromptBuilder.buildWithContext(
-    { questions: questionsContent },
-    [transcriptSection]
-  );
+  const transcriptSection = renderSection(createTranscriptSection(transcriptText, format));
+  return `${transcriptSection}
+
+${questionsSection}
+
+${responseOptions}`;
 }
 async function fetchImageAsBase64(imageUrl, imageDownloadOptions) {
   "use step";
@@ -1257,8 +1298,8 @@ async function askQuestions(assetId, questions, options) {
     cleanTranscript,
     shouldSign: policy === "signed"
   })).transcriptText : "";
-  const userPrompt = buildUserPrompt(questions, transcriptText, cleanTranscript);
-  const systemPrompt = buildSystemPrompt(normalizedAnswerOptions);
+  const userPrompt = buildUserPrompt(questions, allowedAnswers, transcriptText, cleanTranscript);
+  const systemPrompt = SYSTEM_PROMPT;
   const imageUrl = await getStoryboardUrl(
     playbackId,
     storyboardWidth,
@@ -2085,7 +2126,7 @@ async function getThumbnailUrls(playbackId, duration, options = {}) {
     }
     timestamps = newTimestamps;
   }
-  const baseUrl = `https://image.mux.com/${playbackId}/thumbnail.png`;
+  const baseUrl = getMuxThumbnailBaseUrl(playbackId);
   const urlPromises = timestamps.map(async (time) => {
     if (shouldSign) {
       return signUrl(baseUrl, playbackId, "thumbnail", { time, width }, credentials);
@@ -2097,7 +2138,7 @@ async function getThumbnailUrls(playbackId, duration, options = {}) {
 
 // src/workflows/moderation.ts
 var DEFAULT_THRESHOLDS = {
-  sexual: 0.7,
+  sexual: 0.8,
   violence: 0.8
 };
 var DEFAULT_PROVIDER2 = "openai";
@@ -2362,7 +2403,7 @@ async function requestHiveModeration(imageUrls, maxConcurrent = 5, submissionMod
 async function getThumbnailUrlsFromTimestamps(playbackId, timestampsMs, options) {
   "use step";
   const { width, shouldSign, credentials } = options;
-  const baseUrl = `https://image.mux.com/${playbackId}/thumbnail.png`;
+  const baseUrl = getMuxThumbnailBaseUrl(playbackId);
   const urlPromises = timestampsMs.map(async (tsMs) => {
     const time = Number((tsMs / 1e3).toFixed(2));
     if (shouldSign) {
@@ -2539,96 +2580,106 @@ var TONE_INSTRUCTIONS = {
   playful: "Channel your inner diva! Answer with maximum sass, wit, and playful attitude. Don't hold back - be cheeky, clever, and delightfully snarky. Make it pop!",
   professional: "Provide a professional, executive-level analysis suitable for business reporting."
 };
-var summarizationPromptBuilder = createPromptBuilder({
-  template: {
-    task: {
-      tag: "task",
-      content: "Analyze the storyboard frames and generate metadata that captures the essence of the video content."
-    },
-    title: {
-      tag: "title_requirements",
-      content: dedent4`
-        A short, compelling headline that immediately communicates the subject or action.
-        Aim for brevity - typically under 10 words. Think of how a news headline or video card title would read.
-        Start with the primary subject, action, or topic - never begin with "A video of" or similar phrasing.
-        Use active, specific language.`
-    },
-    description: {
-      tag: "description_requirements",
-      content: dedent4`
-        A concise summary (2-4 sentences) that describes what happens across the video.
-        Cover the main subjects, actions, setting, and any notable progression visible across frames.
-        Write in present tense. Be specific about observable details rather than making assumptions.
-        If the transcript provides dialogue or narration, incorporate key points but prioritize visual content.`
-    },
-    keywords: {
-      tag: "keywords_requirements",
-      content: dedent4`
-        Specific, searchable terms (up to ${SUMMARY_KEYWORD_LIMIT}) that capture:
-        - Primary subjects (people, animals, objects)
-        - Actions and activities being performed
-        - Setting and environment
-        - Notable objects or tools
-        - Style or genre (if applicable)
-        Prefer concrete nouns and action verbs over abstract concepts.
-        Use lowercase. Avoid redundant or overly generic terms like "video" or "content".`
-    },
-    qualityGuidelines: {
-      tag: "quality_guidelines",
-      content: dedent4`
-        - Examine all frames to understand the full context and progression
-        - Be precise: "golden retriever" is better than "dog" when identifiable
-        - Capture the narrative: what begins, develops, and concludes
-        - Balance brevity with informativeness`
-    }
-  },
-  sectionOrder: ["task", "title", "description", "keywords", "qualityGuidelines"]
-});
-var audioOnlyPromptBuilder = createPromptBuilder({
-  template: {
-    task: {
-      tag: "task",
-      content: "Analyze the transcript and generate metadata that captures the essence of the audio content."
-    },
-    title: {
-      tag: "title_requirements",
-      content: dedent4`
-        A short, compelling headline that immediately communicates the subject or topic.
-        Aim for brevity - typically under 10 words. Think of how a podcast title or audio description would read.
-        Start with the primary subject, action, or topic - never begin with "An audio of" or similar phrasing.
-        Use active, specific language.`
-    },
-    description: {
-      tag: "description_requirements",
-      content: dedent4`
-        A concise summary (2-4 sentences) that describes the audio content.
-        Cover the main topics, speakers, themes, and any notable progression in the discussion or narration.
-        Write in present tense. Be specific about what is discussed or presented rather than making assumptions.
-        Focus on the spoken content and any key insights, dialogue, or narrative elements.`
+function createSummarizationBuilder({ titleLength, descriptionLength, tagCount } = {}) {
+  const titleBrevity = titleLength != null ? `Aim for approximately ${titleLength} characters.` : "Aim for brevity - typically under 10 words.";
+  const descConstraint = descriptionLength != null ? `approximately ${descriptionLength} characters` : "2-4 sentences";
+  const keywordLimit = tagCount ?? SUMMARY_KEYWORD_LIMIT;
+  return createPromptBuilder({
+    template: {
+      task: {
+        tag: "task",
+        content: "Analyze the storyboard frames and generate metadata that captures the essence of the video content."
+      },
+      title: {
+        tag: "title_requirements",
+        content: dedent4`
+          A short, compelling headline that immediately communicates the subject or action.
+          ${titleBrevity} Think of how a news headline or video card title would read.
+          Start with the primary subject, action, or topic - never begin with "A video of" or similar phrasing.
+          Use active, specific language.`
+      },
+      description: {
+        tag: "description_requirements",
+        content: dedent4`
+          A concise summary (${descConstraint}) that describes what happens across the video.
+          Cover the main subjects, actions, setting, and any notable progression visible across frames.
+          Write in present tense. Be specific about observable details rather than making assumptions.
+          If the transcript provides dialogue or narration, incorporate key points but prioritize visual content.`
+      },
+      keywords: {
+        tag: "keywords_requirements",
+        content: dedent4`
+          Specific, searchable terms (up to ${keywordLimit}) that capture:
+          - Primary subjects (people, animals, objects)
+          - Actions and activities being performed
+          - Setting and environment
+          - Notable objects or tools
+          - Style or genre (if applicable)
+          Prefer concrete nouns and action verbs over abstract concepts.
+          Use lowercase. Avoid redundant or overly generic terms like "video" or "content".`
+      },
+      qualityGuidelines: {
+        tag: "quality_guidelines",
+        content: dedent4`
+          - Examine all frames to understand the full context and progression
+          - Be precise: "golden retriever" is better than "dog" when identifiable
+          - Capture the narrative: what begins, develops, and concludes
+          - Balance brevity with informativeness`
+      }
     },
-    keywords: {
-      tag: "keywords_requirements",
-      content: dedent4`
-        Specific, searchable terms (up to ${SUMMARY_KEYWORD_LIMIT}) that capture:
-        - Primary topics and themes
-        - Speakers or presenters (if named)
-        - Key concepts and terminology
-        - Content type (interview, lecture, music, etc.)
-        - Genre or style (if applicable)
-        Prefer concrete nouns and relevant terms over abstract concepts.
-        Use lowercase. Avoid redundant or overly generic terms like "audio" or "content".`
+    sectionOrder: ["task", "title", "description", "keywords", "qualityGuidelines"]
+  });
+}
+function createAudioOnlyBuilder({ titleLength, descriptionLength, tagCount } = {}) {
+  const titleBrevity = titleLength != null ? `Aim for approximately ${titleLength} characters.` : "Aim for brevity - typically under 10 words.";
+  const descConstraint = descriptionLength != null ? `approximately ${descriptionLength} characters` : "2-4 sentences";
+  const keywordLimit = tagCount ?? SUMMARY_KEYWORD_LIMIT;
+  return createPromptBuilder({
+    template: {
+      task: {
+        tag: "task",
+        content: "Analyze the transcript and generate metadata that captures the essence of the audio content."
+      },
+      title: {
+        tag: "title_requirements",
+        content: dedent4`
+          A short, compelling headline that immediately communicates the subject or topic.
+          ${titleBrevity} Think of how a podcast title or audio description would read.
+          Start with the primary subject, action, or topic - never begin with "An audio of" or similar phrasing.
+          Use active, specific language.`
+      },
+      description: {
+        tag: "description_requirements",
+        content: dedent4`
+          A concise summary (${descConstraint}) that describes the audio content.
+          Cover the main topics, speakers, themes, and any notable progression in the discussion or narration.
+          Write in present tense. Be specific about what is discussed or presented rather than making assumptions.
+          Focus on the spoken content and any key insights, dialogue, or narrative elements.`
+      },
+      keywords: {
+        tag: "keywords_requirements",
+        content: dedent4`
+          Specific, searchable terms (up to ${keywordLimit}) that capture:
+          - Primary topics and themes
+          - Speakers or presenters (if named)
+          - Key concepts and terminology
+          - Content type (interview, lecture, music, etc.)
+          - Genre or style (if applicable)
+          Prefer concrete nouns and relevant terms over abstract concepts.
+          Use lowercase. Avoid redundant or overly generic terms like "audio" or "content".`
+      },
+      qualityGuidelines: {
+        tag: "quality_guidelines",
+        content: dedent4`
+          - Analyze the full transcript to understand context and themes
+          - Be precise: use specific terminology when mentioned
+          - Capture the narrative: what is introduced, discussed, and concluded
+          - Balance brevity with informativeness`
+      }
     },
-    qualityGuidelines: {
-      tag: "quality_guidelines",
-      content: dedent4`
-        - Analyze the full transcript to understand context and themes
-        - Be precise: use specific terminology when mentioned
-        - Capture the narrative: what is introduced, discussed, and concluded
-        - Balance brevity with informativeness`
-    }
-  },
-  sectionOrder: ["task", "title", "description", "keywords", "qualityGuidelines"]
-});
+    sectionOrder: ["task", "title", "description", "keywords", "qualityGuidelines"]
+  });
+}
 var SYSTEM_PROMPT3 = dedent4`
   <role>
     You are a video content analyst specializing in storyboard interpretation and multimodal analysis.
@@ -2743,14 +2794,18 @@ function buildUserPrompt4({
   transcriptText,
   isCleanTranscript = true,
   promptOverrides,
-  isAudioOnly = false
+  isAudioOnly = false,
+  titleLength,
+  descriptionLength,
+  tagCount
 }) {
   const contextSections = [createToneSection(TONE_INSTRUCTIONS[tone])];
   if (transcriptText) {
     const format = isCleanTranscript ? "plain text" : "WebVTT";
     contextSections.push(createTranscriptSection(transcriptText, format));
   }
-  const promptBuilder = isAudioOnly ? audioOnlyPromptBuilder : summarizationPromptBuilder;
+  const constraints = { titleLength, descriptionLength, tagCount };
+  const promptBuilder = isAudioOnly ? createAudioOnlyBuilder(constraints) : createSummarizationBuilder(constraints);
   return promptBuilder.buildWithContext(promptOverrides, contextSections);
 }
 async function analyzeStoryboard2(imageDataUrl, provider, modelId, userPrompt, systemPrompt, credentials) {
@@ -2820,7 +2875,7 @@ async function analyzeAudioOnly(provider, modelId, userPrompt, systemPrompt, cre
     }
   };
 }
-function normalizeKeywords(keywords) {
+function normalizeKeywords(keywords, limit = SUMMARY_KEYWORD_LIMIT) {
   if (!Array.isArray(keywords) || keywords.length === 0) {
     return [];
   }
@@ -2837,7 +2892,7 @@ function normalizeKeywords(keywords) {
     }
     uniqueLowercase.add(lower);
     normalized.push(trimmed);
-    if (normalized.length === SUMMARY_KEYWORD_LIMIT) {
+    if (normalized.length === limit) {
       break;
     }
   }
@@ -2854,7 +2909,10 @@ async function getSummaryAndTags(assetId, options) {
     imageSubmissionMode = "url",
     imageDownloadOptions,
     promptOverrides,
-    credentials
+    credentials,
+    titleLength,
+    descriptionLength,
+    tagCount
   } = options ?? {};
   if (!VALID_TONES.includes(tone)) {
     throw new Error(
@@ -2892,7 +2950,10 @@ async function getSummaryAndTags(assetId, options) {
     transcriptText,
     isCleanTranscript: cleanTranscript,
     promptOverrides,
-    isAudioOnly
+    isAudioOnly,
+    titleLength,
+    descriptionLength,
+    tagCount
   });
   let analysisResponse;
   let imageUrl;
@@ -2949,7 +3010,7 @@ async function getSummaryAndTags(assetId, options) {
     assetId,
     title: analysisResponse.result.title,
     description: analysisResponse.result.description,
-    tags: normalizeKeywords(analysisResponse.result.keywords),
+    tags: normalizeKeywords(analysisResponse.result.keywords, tagCount ?? SUMMARY_KEYWORD_LIMIT),
     storyboardUrl: imageUrl,
     // undefined for audio-only assets
     usage: {
@@ -3781,6 +3842,7 @@ import { z as z6 } from "zod";
 var translationSchema = z6.object({
   translation: z6.string()
 });
+var SYSTEM_PROMPT4 = 'You are a subtitle translation expert. Translate VTT subtitle files to the target language specified by the user. Preserve all timestamps and VTT formatting exactly as they appear. Return JSON with a single key "translation" containing the translated VTT content.';
 async function fetchVttFromMux(vttUrl) {
   "use step";
   const vttResponse = await fetch(vttUrl);
@@ -3803,9 +3865,13 @@ async function translateVttWithAI({
     model,
     output: Output5.object({ schema: translationSchema }),
     messages: [
+      {
+        role: "system",
+        content: SYSTEM_PROMPT4
+      },
       {
         role: "user",
-        content: `Translate the following VTT subtitle file from ${fromLanguageCode} to ${toLanguageCode}. Preserve all timestamps and VTT formatting exactly as they appear. Return JSON with a single key "translation" containing the translated VTT.
+        content: `Translate from ${fromLanguageCode} to ${toLanguageCode}:
 
 ${vttContent}`
       }