@speakableio/core 1.0.64 → 1.0.66

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;
@@ -1699,6 +1700,10 @@ declare function getTranscriptCycle(args: {
     audioUrl: string;
     language: string;
     prompt?: string;
+    cleanHallucinations?: boolean;
+    options?: {
+        modelOrder?: TranscriptModel[];
+    };
 }): Promise<{
     transcript: string;
     success: boolean;
@@ -2967,10 +2972,11 @@ 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;
+        cleanHallucinations?: boolean;
     }, unknown>;
 };
 declare function useSpeakableTranscriptCycle(): {
@@ -2981,6 +2987,10 @@ declare function useSpeakableTranscriptCycle(): {
         audioUrl: string;
         language: string;
         prompt: string;
+        cleanHallucinations?: boolean;
+        options?: {
+            modelOrder?: TranscriptModel[];
+        };
     }, unknown>;
 };
 
@@ -3563,4 +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

@@ -2217,6 +2217,10 @@ function detectTranscriptHallucination(transcript) {
     return false;
   }
   const text = transcript.trim();
+  const wordCount = text.split(/\s+/).filter(Boolean).length;
+  if (text.length < 120 || wordCount < 20) {
+    return false;
+  }
   const shortRepeats = detectShortRepeats(text);
   if (shortRepeats) return true;
   const phraseRepeats = detectPhraseRepeats(text);
@@ -2232,14 +2236,14 @@ function detectShortRepeats(text) {
   for (let i = 1; i < words.length; i++) {
     if (words[i] === words[i - 1]) {
       repeatCount++;
-      if (repeatCount >= 3) return true;
+      if (repeatCount >= 4) return true;
     } else {
       repeatCount = 1;
     }
   }
   const uniqueWords = new Set(words);
   const repetitionRatio = words.length / uniqueWords.size;
-  if (words.length >= 10 && uniqueWords.size <= 3 && repetitionRatio >= 3) {
+  if (words.length >= 12 && uniqueWords.size <= 2 && repetitionRatio >= 5) {
     return true;
   }
   return false;
@@ -2256,12 +2260,12 @@ function detectPhraseRepeats(text) {
         break;
       }
     }
-    if (consecutiveRepeats >= 2) {
+    if (consecutiveRepeats >= 3) {
       return true;
     }
   }
   const uniqueSentences = new Set(sentences);
-  if (sentences.length >= 3 && uniqueSentences.size === 1) {
+  if (sentences.length >= 4 && uniqueSentences.size === 1) {
     return true;
   }
   return false;
@@ -2278,7 +2282,7 @@ function isSimilarSentence(s1, s2) {
   const set2 = new Set(words2);
   const intersection = new Set([...set1].filter((w) => set2.has(w)));
   const similarity = intersection.size * 2 / (set1.size + set2.size);
-  return similarity >= 0.8;
+  return similarity >= 0.9;
 }
 function detectCyclicPattern(text) {
   const normalized = text.toLowerCase().replace(/\s+/g, " ").trim();
@@ -2298,7 +2302,7 @@ function detectCyclicPattern(text) {
         break;
       }
     }
-    if (matchCount >= 3) {
+    if (matchCount >= 4) {
       return true;
     }
   }
@@ -2314,12 +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");
-  console.log("Getting transcript from", model);
+  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({
@@ -2349,7 +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) {
@@ -2372,15 +2390,27 @@ async function getTranscript(model, args, cleanHallucinations = true) {
   return null;
 }
 async function getTranscriptCycle(args) {
-  const models = ["whisper-3", "whisper", "gemini", "assemblyai"];
+  var _a, _b, _c;
+  const models = (_b = (_a = args.options) == null ? void 0 : _a.modelOrder) != null ? _b : [
+    "gpt-4o",
+    "whisper",
+    "whisper-3",
+    "gemini",
+    "assemblyai"
+  ];
   let transcript = "";
   let lastError = null;
   for (const model of models) {
     try {
-      console.log("Getting transcript from", model);
-      const transcriptResult = await getTranscript(model, args, false);
+      console.log(
+        "Getting transcript from",
+        model,
+        " cleanHallucinations",
+        args.cleanHallucinations
+      );
+      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;
@@ -3137,9 +3167,10 @@ function useSpeakableTranscript() {
       model,
       audioUrl,
       language,
-      prompt
+      prompt,
+      cleanHallucinations = true
     }) => {
-      return getTranscript(model, { audioUrl, language, prompt });
+      return getTranscript(model, { audioUrl, language, prompt }, cleanHallucinations);
     },
     retry: false
   });
@@ -3150,7 +3181,11 @@ function useSpeakableTranscript() {
 function useSpeakableTranscriptCycle() {
   const mutation = useMutation3({
     mutationFn: async (args) => {
-      return getTranscriptCycle(args);
+      return getTranscriptCycle({
+        ...args,
+        cleanHallucinations: args.cleanHallucinations,
+        options: args.options
+      });
     },
     retry: false
   });