emillindfors/transcriber-backend

Transcriber

Transcriber is a web app using Google speech-to-text API for transcribing audio files. Transcoding, transcription and database is handled by Cloud functions and Firebase, while React JS is used for the web frontend.

How to use this API

1. Generate upload url:

curl -H "Content-Type: audio/*" \ 
  "https://<region>-<project-id>.cloudfunctions.net/getUploadUrl"

2. Upload an audio file

curl -X PUT --data-binary @$FILE_PATH \
    -H "Content-Type: audio/*" \
    "<upload-url>"

3. Transcribe: curl https://<region>-<project-id>.cloudfunctions.net/swagger
4. Export transcriptions: curl https://<region>-<project-id>.cloudfunctions.net/swagger

Tech overview

• Google Cloud Storage
• Cloud Functions for Firebase
• Cloud Speech-to-text
• Cloud Firestore

Set up

• Create a Firebase project
• Turn on the Firestore database and Storage. -- TODO how to do this? Prefer cli/.sh -- TODO enable security for generating uploadUrl
• Copy [.firebaserc_sample] to .firebaserc
• Edit .firebaserc with the name of your Firebase project.
• Install the Firebase CLI: npm install -g firebase-tools

Firebase Storage

• Use the default bucket in Firebase Storage, or create a new one TODO how to define default bucket in config

Deploy environment variables

set up environment variables with the name of the bucket you just created, along with your Google Analytics account ID:

> firebase functions:config:set \
bucket.name="name-of-bucket" \
webserver.domainname="https://www.example.com"

Enable Google Speech

• Enable the Google Speech API.

Config

cd functions firebase functions:config:set bucket.name=transcribe-baardl

Testing

Create a .env file in the test folder with the following attributes:

FIREBASE_DATABASE_URL = https:...
FIREBASE_UPLOADS_BUCKET = name-of-uploads-bucket
FIREBASE_TRANSCODED_BUCKET = name-of-transcoded-bucket

Deployment

Configuration

> firebase functions:config:set \
analytics.account_id="UA-XXXXXX-XX" \
bucket.name="name-of-bucket" \
sendgrid.apikey="api key" \
sendgrid.email="[email protected]" \
sendgrid.name="Your name" \
webserver.domainname="https://www.example.com"

Deploy code

TODO From which directory?

cd functions
npm run deploy

or?

firebase deploy

Verifications

TODO eg curl https://<region>-<project-id>.cloudfunctions.net/health

API Documentation

TODO Use Swagger curl https://<region>-<project-id>.cloudfunctions.net/swagger

Google Analytics

Exceptions are logged.

Transcription

Custom dimensions

• cd1: Language codes
• cd2: Original MIME type
• cd3: Industry NAICS code of audio
• cd4: Interaction type
• cd5: Microphone distance
• cd6: Original media type
• cd7: Recording device name
• cd8: Recording device type

Custom metrics

• cm1: Number of audio topic words
• cm2: Number of speech contexts phrases
• cm3: Audio duration
• cm4: Number of words
• cm5: Transcoding duration
• cm6: Transcribing duration
• cm7: Saving duration
• cm8: Process duration (transcoding + transcribing + saving)
• cm9: Confidence

Events

Category	Action	Label	Value
transcription	transcoded	transcript id
transcription	transcribed	transcript id
transcription	saved	transcript id
transcription	done	transcript id	audio duration
api	authorization	idtoken	token
api	authorization	customtoken	token

User timings

• transcription → transcoding
• transcription → transcribing
• transcription → saving

Export

Events

Category	Action	Label
transcript	exported	[type]
transcript	deleted	step
email	transcription done	transcript id