From 02ba753375ac559745f34585ce3aed3c560da708 Mon Sep 17 00:00:00 2001
From: Hongbo Wu <hongbo@omnivore.app>
Date: Wed, 17 Aug 2022 11:24:53 +0800
Subject: [PATCH] Upload speech marks file to gcs too

---
 packages/api/src/entity/speech.ts          |  2 +-
 packages/api/src/routers/article_router.ts | 41 +++++++++++++++++-----
 packages/api/src/utils/textToSpeech.ts     | 16 +++++++--
 packages/db/migrations/0093.do.speech.sql  |  2 +-
 4 files changed, 48 insertions(+), 13 deletions(-)

diff --git a/packages/api/src/entity/speech.ts b/packages/api/src/entity/speech.ts
index 0f61d7dd7..ac64f7e71 100644
--- a/packages/api/src/entity/speech.ts
+++ b/packages/api/src/entity/speech.ts
@@ -25,7 +25,7 @@ export class Speech {
   audioUrl!: string
 
   @Column('text')
-  speechMarks!: string
+  speechMarksUrl!: string
 
   @Column('text')
   voice!: string
diff --git a/packages/api/src/routers/article_router.ts b/packages/api/src/routers/article_router.ts
index d9baf737c..85ccf9b0a 100644
--- a/packages/api/src/routers/article_router.ts
+++ b/packages/api/src/routers/article_router.ts
@@ -72,10 +72,11 @@ export function articleRouter() {
   })
 
   router.get(
-    '/:id/mp3',
+    '/:id/:outputFormat',
     cors<express.Request>(corsConfig),
     async (req, res) => {
       const id = req.params.id
+      const outputFormat = req.params.outputFormat
       const token = req.cookies?.auth || req.headers?.authorization
       if (!token || !jwt.verify(token, env.server.jwtSecret)) {
         return res.status(401).send({ errorCode: 'UNAUTHORIZED' })
@@ -83,15 +84,27 @@ export function articleRouter() {
       const { uid } = jwt.decode(token) as Claims
 
       const startTime = Date.now()
-      logger.info('Get article speech in mp3 format', {
+      logger.info(`Get article speech in ${outputFormat} format`, {
         params: req.params,
         labels: {
           userId: uid,
           source: 'GetArticleSpeechMp3',
           articleId: id,
+          outputFormat,
         },
       })
 
+      const existingSpeech = await getRepository(Speech).findOneBy({
+        elasticPageId: id,
+      })
+      if (existingSpeech) {
+        logger.info('Found existing speech', {
+          audioUrl: existingSpeech.audioUrl,
+          speechMarksUrl: existingSpeech.speechMarksUrl,
+        })
+        return res.redirect(redirectUrl(existingSpeech, outputFormat))
+      }
+
       logger.debug('Text to speech request', { articleId: id })
       const userPersonalization = await getRepository(
         UserPersonalization
@@ -112,27 +125,39 @@ export function articleRouter() {
         return res.status(200).send('Page has no text')
       }
 
-      const speech = await synthesizeTextToSpeech({
+      const speechOutput = await synthesizeTextToSpeech({
         id,
         text,
         languageCode: page.language,
         voice: userPersonalization.speechVoice,
       })
 
-      await getRepository(Speech).save({
+      const speech = await getRepository(Speech).save({
         elasticPageId: id,
-        audioUrl: speech.audioUrl,
-        speechMarks: JSON.stringify(speech.speechMarks),
+        audioUrl: speechOutput.audioUrl,
+        speechMarksUrl: speechOutput.speechMarksUrl,
         user: { id: uid },
       })
 
-      logger.info('Found speech mp3', {
+      logger.info('Created speech', {
         audioUrl: speech.audioUrl,
+        speechMarksUrl: speech.speechMarksUrl,
         duration: Date.now() - startTime,
       })
-      res.redirect(speech.audioUrl)
+      res.redirect(redirectUrl(speech, outputFormat))
     }
   )
 
   return router
 }
+
+const redirectUrl = (speech: Speech, outputFormat: string) => {
+  switch (outputFormat) {
+    case 'mp3':
+      return speech.audioUrl
+    case 'json':
+      return speech.speechMarksUrl
+    default:
+      return speech.audioUrl
+  }
+}
diff --git a/packages/api/src/utils/textToSpeech.ts b/packages/api/src/utils/textToSpeech.ts
index 53aefe881..7f1a1d45c 100644
--- a/packages/api/src/utils/textToSpeech.ts
+++ b/packages/api/src/utils/textToSpeech.ts
@@ -1,5 +1,5 @@
 import { buildLogger } from './logger'
-import { createGCSFile, getFilePublicUrl } from './uploads'
+import { createGCSFile, getFilePublicUrl, uploadToBucket } from './uploads'
 import {
   CancellationDetails,
   CancellationReason,
@@ -21,7 +21,7 @@ export interface TextToSpeechInput {
 
 export interface TextToSpeechOutput {
   audioUrl: string
-  speechMarks: SpeechMark[]
+  speechMarksUrl: string
 }
 
 export interface SpeechMark {
@@ -142,9 +142,19 @@ export const synthesizeTextToSpeech = async (
 
   logger.debug(`audio file: ${audioFile}`)
 
+  // upload Speech Marks file to GCS
+  const speechMarksFile = `speech/${input.id}.json`
+  await uploadToBucket(
+    speechMarksFile,
+    Buffer.from(JSON.stringify(speechMarks)),
+    {
+      public: true,
+    }
+  )
+
   return {
     audioUrl: getFilePublicUrl(audioFile),
-    speechMarks,
+    speechMarksUrl: getFilePublicUrl(speechMarksFile),
   }
 }
 
diff --git a/packages/db/migrations/0093.do.speech.sql b/packages/db/migrations/0093.do.speech.sql
index 39a50aadf..fcda74145 100755
--- a/packages/db/migrations/0093.do.speech.sql
+++ b/packages/db/migrations/0093.do.speech.sql
@@ -10,7 +10,7 @@ CREATE TABLE omnivore.speech (
     elastic_page_id TEXT NOT NULL,
     voice text,
     audio_url text NOT NULL,
-    speech_marks text NOT NULL,
+    speech_marks_url text NOT NULL,
     created_at timestamptz NOT NULL DEFAULT current_timestamp,
     updated_at timestamptz NOT NULL DEFAULT current_timestamp
 );