Enqueue text to speech tasks

packages/api/src/routers/text_to_speech.ts

@@ -9,9 +9,10 @@ import { getPageById } from '../elastic/pages'
 import { Speech, SpeechState } from '../entity/speech'
 import { buildLogger } from '../utils/logger'
 import { getClaimsByToken } from '../utils/auth'
-import { shouldSynthesize, synthesize } from '../services/speech'
+import { shouldSynthesize } from '../services/speech'
 import { readPushSubscription } from '../datalayer/pubsub'
 import { AppDataSource } from '../server'
+import { enqueueTextToSpeech } from '../utils/createTask'
 
 const logger = buildLogger('app.dispatch')
 
@@ -62,8 +63,16 @@ export function textToSpeechRouter() {
           state: SpeechState.INITIALIZED,
           voice: 'en-US-JennyNeural',
         })
-        await synthesize(page, speech)
-        logger.info('page synthesized')
+        // enqueue a task to convert text to speech
+        const taskName = await enqueueTextToSpeech({
+          userId,
+          speechId: speech.id,
+          text: page.content,
+          voice: speech.voice,
+          priority: 'low',
+        })
+        logger.info('Start Text to speech task', { taskName })
+        return res.status(202).send('Text to speech task started')
       }
 
       res.status(200).send('Page should not synthesize')

packages/api/src/services/speech.ts

@@ -1,16 +1,6 @@
-import { getRepository } from '../entity/utils'
-import { Speech, SpeechState } from '../entity/speech'
 import { searchPages } from '../elastic/pages'
 import { Page, PageType } from '../elastic/types'
 import { SortBy, SortOrder } from '../utils/search'
-import { synthesizeTextToSpeech } from '../utils/textToSpeech'
-
-export const setSpeechFailure = async (id: string) => {
-  // update state
-  await getRepository(Speech).update(id, {
-    state: SpeechState.FAILED,
-  })
-}
 
 /*
  * We should not synthesize the page when:
@@ -54,38 +44,3 @@ export const shouldSynthesize = async (
     page.savedAt < recentListenedPage[0].listenedAt
   )
 }
-
-export const synthesize = async (page: Page, speech: Speech): Promise<void> => {
-  try {
-    if (page.pageType === PageType.File || !page.content) {
-      // we don't synthesize files for now
-      return
-    }
-
-    console.log('Start synthesizing', { pageId: page.id, speechId: speech.id })
-    const startTime = Date.now()
-    const speechOutput = await synthesizeTextToSpeech({
-      id: speech.id,
-      text: page.content,
-      languageCode: page.language,
-      voice: speech.voice,
-      textType: 'ssml',
-    })
-    console.log('Synthesized article', {
-      audioFileName: speechOutput.audioFileName,
-      speechMarksFileName: speechOutput.speechMarksFileName,
-      duration: Date.now() - startTime,
-    })
-
-    // set state to completed
-    await getRepository(Speech).update(speech.id, {
-      audioFileName: speechOutput.audioFileName,
-      speechMarksFileName: speechOutput.speechMarksFileName,
-      state: SpeechState.COMPLETED,
-    })
-  } catch (error) {
-    console.log('Error synthesize article', error)
-    await setSpeechFailure(speech.id)
-    throw error
-  }
-}