@speakableio/core 1.0.65 → 1.0.67

package/dist/index.web.d.mts

@@ -1690,7 +1690,8 @@ declare function SpeakableProvider({ user, children, queryClient, permissions, f
 }): react_jsx_runtime.JSX.Element | null;
 declare function useSpeakableApi(): FsContext;
 
-declare function getTranscript(model: 'gemini' | 'assemblyai' | 'whisper' | 'whisper-3', args: {
+type TranscriptModel = 'gemini' | 'assemblyai' | 'whisper' | 'whisper-3' | 'gpt-4o';
+declare function getTranscript(model: TranscriptModel, args: {
     language: string;
     audioUrl: string;
     prompt?: string;
@@ -1700,6 +1701,9 @@ declare function getTranscriptCycle(args: {
     language: string;
     prompt?: string;
     cleanHallucinations?: boolean;
+    options?: {
+        modelOrder?: TranscriptModel[];
+    };
 }): Promise<{
     transcript: string;
     success: boolean;
@@ -2968,7 +2972,7 @@ declare const useOrganizationAccess: () => {
 
 declare function useSpeakableTranscript(): {
     mutation: _tanstack_react_query.UseMutationResult<string | null, Error, {
-        model: "gemini" | "assemblyai" | "whisper" | "whisper-3";
+        model: TranscriptModel;
         audioUrl: string;
         language: string;
         prompt?: string;
@@ -2984,6 +2988,9 @@ declare function useSpeakableTranscriptCycle(): {
         language: string;
         prompt: string;
         cleanHallucinations?: boolean;
+        options?: {
+            modelOrder?: TranscriptModel[];
+        };
     }, unknown>;
 };
 
@@ -3566,4 +3573,4 @@ declare const createFsClientWeb: ({ db, httpsCallable, logEvent }: FsClientParam
     };
 };
 
-export { ActivityPageType, type Assignment, type AssignmentAnalyticsType$1 as AssignmentAnalyticsType, type AssignmentWithId, BASE_MULTIPLE_CHOICE_FIELD_VALUES, BASE_REPEAT_FIELD_VALUES, BASE_RESPOND_FIELD_VALUES, CONVERSATION_PAGE_ACTIVITY_TYPES, ConversationPageMode, type CreditContract, FeedbackTypesCard, FsCtx, type InstitutionSubscription, LENIENCY_OPTIONS, LeniencyCard, MULTIPLE_CHOICE_PAGE_ACTIVITY_TYPES, type MediaPageActivity, type Organization, type OrganizationAccess, type PageActivity, type PageActivityWithId, type PageScore, REPEAT_PAGE_ACTIVITY_TYPES, RESPOND_AUDIO_PAGE_ACTIVITY_TYPES, RESPOND_PAGE_ACTIVITY_TYPES, RESPOND_WRITE_PAGE_ACTIVITY_TYPES, type RefsCardsFiresotre, type RefsSetsFirestore, SPEAKABLE_ANALYTICS, SPEAKABLE_NOTIFICATIONS, STUDENT_LEVELS_OPTIONS, type Score, type ScoreWithId, type Set, type SetWithId, type SpeakableNotificationType, SpeakableNotificationTypes, SpeakableProvider, VerificationCardStatus, assignmentQueryKeys, cardsQueryKeys, checkIsConversationPage, checkIsMCPage, checkIsMediaPage, checkIsRepeatPage, checkIsRespondAudioPage, checkIsRespondPage, checkIsRespondWrittenPage, checkIsShortAnswerPage, checkTypePageActivity, cleanString, createAssignmentRepo, createCardRepo, createFsClientWeb as createFsClient, createSetRepo, creditQueryKeys, debounce, getCardFromCache, getLabelPage, getPageMediaData, getPagePrompt, getPhraseLength, getRespondCardTool, getSetFromCache, getSingleMediaPageData, getTotalCompletedCards, getTranscript, getTranscriptCycle, getWordHash, purify, refsCardsFiresotre, refsSetsFirestore, scoreQueryKeys, setsQueryKeys, updateCardInCache, updateSetInCache, useActivity, useActivityFeedbackAccess, useAssignment, useBaseOpenAI, useCards, useClearScore, useClearScoreV2, useCreateCard, useCreateCards, useCreateNotification, useGetCard, useOrganizationAccess, useScore, useSet, useSpeakableApi, useSpeakableTranscript, useSpeakableTranscriptCycle, useSubmitAssignmentScore, useSubmitPracticeScore, useUpdateCardScore, useUpdateScore, useUpdateStudentVocab, useUserCredits };
+export { ActivityPageType, type Assignment, type AssignmentAnalyticsType$1 as AssignmentAnalyticsType, type AssignmentWithId, BASE_MULTIPLE_CHOICE_FIELD_VALUES, BASE_REPEAT_FIELD_VALUES, BASE_RESPOND_FIELD_VALUES, CONVERSATION_PAGE_ACTIVITY_TYPES, ConversationPageMode, type CreditContract, FeedbackTypesCard, FsCtx, type InstitutionSubscription, LENIENCY_OPTIONS, LeniencyCard, MULTIPLE_CHOICE_PAGE_ACTIVITY_TYPES, type MediaPageActivity, type Organization, type OrganizationAccess, type PageActivity, type PageActivityWithId, type PageScore, REPEAT_PAGE_ACTIVITY_TYPES, RESPOND_AUDIO_PAGE_ACTIVITY_TYPES, RESPOND_PAGE_ACTIVITY_TYPES, RESPOND_WRITE_PAGE_ACTIVITY_TYPES, type RefsCardsFiresotre, type RefsSetsFirestore, SPEAKABLE_ANALYTICS, SPEAKABLE_NOTIFICATIONS, STUDENT_LEVELS_OPTIONS, type Score, type ScoreWithId, type Set, type SetWithId, type SpeakableNotificationType, SpeakableNotificationTypes, SpeakableProvider, type TranscriptModel, VerificationCardStatus, assignmentQueryKeys, cardsQueryKeys, checkIsConversationPage, checkIsMCPage, checkIsMediaPage, checkIsRepeatPage, checkIsRespondAudioPage, checkIsRespondPage, checkIsRespondWrittenPage, checkIsShortAnswerPage, checkTypePageActivity, cleanString, createAssignmentRepo, createCardRepo, createFsClientWeb as createFsClient, createSetRepo, creditQueryKeys, debounce, getCardFromCache, getLabelPage, getPageMediaData, getPagePrompt, getPhraseLength, getRespondCardTool, getSetFromCache, getSingleMediaPageData, getTotalCompletedCards, getTranscript, getTranscriptCycle, getWordHash, purify, refsCardsFiresotre, refsSetsFirestore, scoreQueryKeys, setsQueryKeys, updateCardInCache, updateSetInCache, useActivity, useActivityFeedbackAccess, useAssignment, useBaseOpenAI, useCards, useClearScore, useClearScoreV2, useCreateCard, useCreateCards, useCreateNotification, useGetCard, useOrganizationAccess, useScore, useSet, useSpeakableApi, useSpeakableTranscript, useSpeakableTranscriptCycle, useSubmitAssignmentScore, useSubmitPracticeScore, useUpdateCardScore, useUpdateScore, useUpdateStudentVocab, useUserCredits };

package/dist/index.web.js

@@ -2318,12 +2318,26 @@ function cleanHallucinatedTranscript(transcript) {
 
 // src/utils/ai/get-transcript.ts
 async function getTranscript(model, args, cleanHallucinations = true) {
-  var _a, _b, _c, _d, _e, _f, _g, _h, _i;
+  var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l;
   const getGeminiTranscript = (_b = (_a = api).httpsCallable) == null ? void 0 : _b.call(_a, "getGeminiTranscript");
   const getAssemblyAITranscript = (_d = (_c = api).httpsCallable) == null ? void 0 : _d.call(_c, "transcribeAssemblyAIAudio");
   const getWhisper3Transcript = (_f = (_e = api).httpsCallable) == null ? void 0 : _f.call(_e, "generateGroqTranscript");
   const getWhisper1Transcript = (_h = (_g = api).httpsCallable) == null ? void 0 : _h.call(_g, "transcribeAudio");
+  const getGPT4oTranscript = (_j = (_i = api).httpsCallable) == null ? void 0 : _j.call(_i, "generateGpt4oTranscript");
   console.log("Getting transcript from", model, " cleanHallucinations", cleanHallucinations);
+  if (model === "gpt-4o") {
+    try {
+      const { data } = await (getGPT4oTranscript == null ? void 0 : getGPT4oTranscript({
+        audioUrl: args.audioUrl,
+        language: args.language,
+        teacherPrompt: (_k = args.prompt) != null ? _k : ""
+      }));
+      return data;
+    } catch (error) {
+      console.error("Error getting transcript from GPT-4o:", error);
+      throw error;
+    }
+  }
   if (model === "whisper-3") {
     try {
       const { data } = await (getWhisper3Transcript == null ? void 0 : getWhisper3Transcript({
@@ -2353,7 +2367,7 @@ async function getTranscript(model, args, cleanHallucinations = true) {
       const { data } = await (getGeminiTranscript == null ? void 0 : getGeminiTranscript({
         audioUrl: args.audioUrl,
         targetLanguage: args.language,
-        prompt: (_i = args.prompt) != null ? _i : ""
+        prompt: (_l = args.prompt) != null ? _l : ""
       }));
       return cleanHallucinations ? cleanHallucinatedTranscript(data.transcript) : data.transcript;
     } catch (error) {
@@ -2376,10 +2390,17 @@ async function getTranscript(model, args, cleanHallucinations = true) {
   return null;
 }
 async function getTranscriptCycle(args) {
-  var _a;
-  const models = ["whisper-3", "whisper", "gemini", "assemblyai"];
+  var _a, _b, _c;
+  const models = (_b = (_a = args.options) == null ? void 0 : _a.modelOrder) != null ? _b : [
+    "whisper",
+    "whisper-3",
+    "gpt-4o",
+    "gemini",
+    "assemblyai"
+  ];
   let transcript = "";
   let lastError = null;
+  console.log("Getting transcript cycle with models", models);
   for (const model of models) {
     try {
       console.log(
@@ -2388,9 +2409,9 @@ async function getTranscriptCycle(args) {
         " cleanHallucinations",
         args.cleanHallucinations
       );
-      const transcriptResult = await getTranscript(model, args, (_a = args.cleanHallucinations) != null ? _a : true);
+      const transcriptResult = await getTranscript(model, args, (_c = args.cleanHallucinations) != null ? _c : true);
       const rawTranscript = transcriptResult || "";
-      transcript = cleanHallucinatedTranscript(rawTranscript);
+      transcript = rawTranscript;
       if (transcript !== "") {
         console.log(`Successfully got transcript from ${model}`);
         break;
@@ -3161,7 +3182,11 @@ function useSpeakableTranscript() {
 function useSpeakableTranscriptCycle() {
   const mutation = useMutation3({
     mutationFn: async (args) => {
-      return getTranscriptCycle({ ...args, cleanHallucinations: args.cleanHallucinations });
+      return getTranscriptCycle({
+        ...args,
+        cleanHallucinations: args.cleanHallucinations,
+        options: args.options
+      });
     },
     retry: false
   });