Create speech while requesting

packages/api/src/routers/article_router.ts

@@ -17,6 +17,10 @@ import { env } from '../env'
 import { Claims } from '../resolvers/types'
 import { getRepository } from '../entity/utils'
 import { Speech } from '../entity/speech'
+import { getPageById } from '../elastic/pages'
+import { parseHTML } from 'linkedom'
+import { synthesizeTextToSpeech } from '../utils/textToSpeech'
+import { UserPersonalization } from '../entity/user_personalization'
 
 const logger = buildLogger('app.dispatch')
 
@@ -87,15 +91,40 @@ export function articleRouter() {
         },
       })
 
-      const speech = await getRepository(Speech).findOneBy({
-        elasticPageId: id,
+      logger.debug('Text to speech request', { articleId: id })
+      const userPersonalization = await getRepository(
+        UserPersonalization
+      ).findOneBy({
         user: { id: uid },
       })
-
-      if (!speech) {
-        return res.status(404).send({ errorCode: 'NOT_FOUND' })
+      if (!userPersonalization) {
+        return res.status(200).send('userPersonalization not found')
       }
 
+      const page = await getPageById(id)
+      if (!page) {
+        return res.status(200).send('Page not found')
+      }
+
+      const text = parseHTML(page.content).document.documentElement.textContent
+      if (!text) {
+        return res.status(200).send('Page has no text')
+      }
+
+      const speech = await synthesizeTextToSpeech({
+        id,
+        text,
+        languageCode: page.language,
+        voice: userPersonalization.speechVoice,
+      })
+
+      await getRepository(Speech).save({
+        elasticPageId: id,
+        audioUrl: speech.audioUrl,
+        speechMarks: JSON.stringify(speech.speechMarks),
+        user: { id: uid },
+      })
+
       logger.info('Found speech mp3', { audioUrl: speech.audioUrl })
       res.redirect(speech.audioUrl)
     }

packages/api/src/routers/svc/speech.ts

@@ -36,7 +36,7 @@ export function speechServiceRouter() {
       return res.status(200).send('Page not found')
     }
 
-    const text = parseHTML(page.content).document.textContent
+    const text = parseHTML(page.content).document.documentElement.textContent
     if (!text) {
       return res.status(200).send('Page has no text')
     }

packages/api/src/server.ts

@@ -120,7 +120,7 @@ export const createApp = (): {
   app.use('/svc/pubsub/integrations', integrationsServiceRouter())
   app.use('/svc/reminders', remindersServiceRouter())
   app.use('/svc/pdf-attachments', pdfAttachmentsRouter())
-  app.use('/svc/speech', speechServiceRouter())
+  app.use('/svc/text-to-speech', speechServiceRouter())
 
   if (env.dev.isLocal) {
     app.use('/local/debug', localDebugRouter())

packages/api/src/utils/createTask.ts

@@ -337,7 +337,15 @@ export const enqueueTextToSpeech = async (
 
   // If there is no Google Cloud Project Id exposed, it means that we are in local environment
   if (env.dev.isLocal || !GOOGLE_CLOUD_PROJECT) {
-    return nanoid()
+    // Calling the handler function directly.
+    setTimeout(() => {
+      axios
+        .post(env.queue.textToSpeechTaskHandlerUrl, payload)
+        .catch((error) => {
+          logger.error(error)
+        })
+    }, 0)
+    return ''
   }
 
   const createdTasks = await createHttpTaskWithToken({