@mux/ai 0.7.3 → 0.7.4

package/dist/index.js

@@ -5,7 +5,7 @@ var __export = (target, all) => {
 };
 
 // package.json
-var version = "0.7.3";
+var version = "0.7.4";
 
 // src/env.ts
 import { z } from "zod";
@@ -48,6 +48,9 @@ var EnvSchema = z.object({
   ),
   MUX_TEST_ASSET_ID_AUDIO_ONLY: optionalString("Mux test asset ID for audio-only assets.", "Mux test asset id for audio-only assets for testing"),
   MUX_TEST_ASSET_ID_VIOLENT_AUDIO_ONLY: optionalString("Mux test asset ID for audio-only assets with violent content.", "Mux test asset id for audio-only assets with violent content for testing"),
+  // Eval config
+  MUX_AI_EVAL_MODEL_SET: optionalString("Eval model selection mode.", "Choose between 'default' (provider defaults only) or 'all' (all configured models)"),
+  MUX_AI_EVAL_MODELS: optionalString("Comma-separated eval model pairs.", "Comma-separated provider:model pairs (e.g. 'openai:gpt-5.1,anthropic:claude-sonnet-4-5,google:gemini-3-flash-preview')"),
   // AI Providers
   OPENAI_API_KEY: optionalString("OpenAI API key for OpenAI-backed workflows.", "OpenAI API key"),
   ANTHROPIC_API_KEY: optionalString("Anthropic API key for Claude-backed workflows.", "Anthropic API key"),
@@ -808,6 +811,81 @@ var DEFAULT_EMBEDDING_MODELS = {
   openai: "text-embedding-3-small",
   google: "gemini-embedding-001"
 };
+var LANGUAGE_MODELS = {
+  openai: ["gpt-5.1", "gpt-5-mini"],
+  anthropic: ["claude-sonnet-4-5"],
+  google: ["gemini-3-flash-preview", "gemini-2.5-flash"]
+};
+function getDefaultEvalModelConfigs() {
+  return Object.entries(DEFAULT_LANGUAGE_MODELS).map(([provider, modelId]) => ({ provider, modelId }));
+}
+function getAllEvalModelConfigs() {
+  return Object.entries(LANGUAGE_MODELS).flatMap(([provider, models]) => models.map((modelId) => ({ provider, modelId })));
+}
+function isSupportedProvider(value) {
+  return value === "openai" || value === "anthropic" || value === "google";
+}
+function parseEvalModelPair(value) {
+  const trimmed = value.trim();
+  const [providerRaw, modelIdRaw] = trimmed.split(":", 2);
+  const provider = providerRaw?.trim();
+  const modelId = modelIdRaw?.trim();
+  if (!provider || !modelId) {
+    throw new Error(
+      `Invalid eval model pair "${value}". Use "provider:model" (example: "openai:gpt-5.1").`
+    );
+  }
+  if (!isSupportedProvider(provider)) {
+    throw new Error(
+      `Unsupported eval provider "${provider}" in "${value}". Supported providers: ${Object.keys(LANGUAGE_MODELS).join(", ")}.`
+    );
+  }
+  const supportedModels = LANGUAGE_MODELS[provider];
+  if (!supportedModels.includes(modelId)) {
+    throw new Error(
+      `Unsupported eval model "${modelId}" for provider "${provider}". Supported models: ${supportedModels.join(", ")}.`
+    );
+  }
+  return {
+    provider,
+    modelId
+  };
+}
+function resolveEvalModelConfigs(options = {}) {
+  const explicitPairs = options.modelPairs?.map((value) => value.trim()).filter(Boolean) ?? [];
+  if (explicitPairs.length > 0) {
+    const dedupedPairs = Array.from(new Set(explicitPairs));
+    return dedupedPairs.map(parseEvalModelPair);
+  }
+  const selection = options.selection ?? "default";
+  if (selection === "all") {
+    return getAllEvalModelConfigs();
+  }
+  return getDefaultEvalModelConfigs();
+}
+function resolveEvalModelConfigsFromEnv(environment = env_default) {
+  const rawSelection = environment.MUX_AI_EVAL_MODEL_SET?.trim();
+  const rawModelPairs = environment.MUX_AI_EVAL_MODELS?.trim();
+  let selection;
+  if (!rawSelection || rawSelection === "default") {
+    selection = "default";
+  } else if (rawSelection === "all") {
+    selection = "all";
+  } else {
+    throw new Error(
+      `Invalid MUX_AI_EVAL_MODEL_SET="${rawSelection}". Expected "default" or "all".`
+    );
+  }
+  let modelPairs;
+  if (rawModelPairs) {
+    modelPairs = rawModelPairs.split(",").map((value) => value.trim()).filter(Boolean);
+  }
+  return resolveEvalModelConfigs({
+    selection,
+    modelPairs
+  });
+}
+var EVAL_MODEL_CONFIGS = resolveEvalModelConfigsFromEnv();
 function resolveLanguageModelConfig(options = {}) {
   const provider = options.provider || "openai";
   const modelId = options.model || DEFAULT_LANGUAGE_MODELS[provider];
@@ -3597,6 +3675,18 @@ function getReadyAudioStaticRendition(asset) {
   );
 }
 var hasReadyAudioStaticRendition = (asset) => Boolean(getReadyAudioStaticRendition(asset));
+function getAudioStaticRenditionStatus(asset) {
+  const files = asset.static_renditions?.files;
+  const audioRendition = files?.find((rendition) => rendition.name === "audio.m4a");
+  if (typeof audioRendition?.status === "string" && audioRendition.status.length > 0) {
+    return audioRendition.status;
+  }
+  const aggregateStatus = asset.static_renditions?.status;
+  if (typeof aggregateStatus === "string" && aggregateStatus.length > 0) {
+    return aggregateStatus;
+  }
+  return asset.static_renditions ? "requested" : "not_requested";
+}
 async function requestStaticRenditionCreation(assetId, credentials) {
   "use step";
   const muxClient = await resolveMuxClient(credentials);
@@ -3642,7 +3732,7 @@ async function waitForAudioStaticRendition({
     if (hasReadyAudioStaticRendition(currentAsset)) {
       return currentAsset;
     }
-    const currentStatus = currentAsset.static_renditions?.status || "unknown";
+    const currentStatus = getAudioStaticRenditionStatus(currentAsset);
     console.warn(
       `\u231B Waiting for static rendition (attempt ${attempt}/${STATIC_RENDITION_MAX_ATTEMPTS}) \u2192 ${currentStatus}`
     );
@@ -3668,6 +3758,7 @@ async function createElevenLabsDubbingJob({
   audioBuffer,
   assetId,
   elevenLabsLangCode,
+  elevenLabsSourceLangCode,
   numSpeakers,
   credentials
 }) {
@@ -3677,8 +3768,14 @@ async function createElevenLabsDubbingJob({
   const formData = new FormData();
   formData.append("file", audioBlob);
   formData.append("target_lang", elevenLabsLangCode);
+  if (elevenLabsSourceLangCode) {
+    formData.append("source_lang", elevenLabsSourceLangCode);
+  }
   formData.append("num_speakers", numSpeakers.toString());
-  formData.append("name", `Mux Asset ${assetId} - auto to ${elevenLabsLangCode}`);
+  formData.append(
+    "name",
+    `Mux Asset ${assetId} - ${elevenLabsSourceLangCode ?? "auto"} to ${elevenLabsLangCode}`
+  );
   const dubbingResponse = await fetch("https://api.elevenlabs.io/v1/dubbing", {
     method: "POST",
     headers: {
@@ -3787,6 +3884,7 @@ async function translateAudio(assetId, toLanguageCode, options = {}) {
   "use workflow";
   const {
     provider = "elevenlabs",
+    fromLanguageCode,
     numSpeakers = 0,
     // 0 = auto-detect
     uploadToMux = true,
@@ -3836,13 +3934,18 @@ async function translateAudio(assetId, toLanguageCode, options = {}) {
   }
   console.warn("\u{1F399}\uFE0F Creating dubbing job in ElevenLabs...");
   const elevenLabsLangCode = toISO639_3(toLanguageCode);
-  console.warn(`\u{1F50D} Creating dubbing job for asset ${assetId} with language code: ${elevenLabsLangCode}`);
+  const normalizedFromLanguageCode = fromLanguageCode?.trim();
+  const elevenLabsSourceLangCode = normalizedFromLanguageCode ? toISO639_3(normalizedFromLanguageCode) : void 0;
+  console.warn(
+    `\u{1F50D} Creating dubbing job for asset ${assetId}: ${elevenLabsSourceLangCode ?? "auto"} -> ${elevenLabsLangCode}`
+  );
   let dubbingId;
   try {
     dubbingId = await createElevenLabsDubbingJob({
       audioBuffer,
       assetId,
       elevenLabsLangCode,
+      elevenLabsSourceLangCode,
       numSpeakers,
       credentials
     });