voice-of-interconnect

Offline First demo app for IBM InterConnect 2017

voice-of-interconnect is a simple HTML/CSS/JavaScript app and uses Hoodie for its backend.

Requirements

You need Node.js v6 or newer. We recommend the latest Node LTS version. For local development, the Hoodie Backend is using PouchDB to persist its data. For production, a CouchDB or Cloudant is recommended

If you use the Speech to Text service and record audio in Chrome, you need ffmpeg to be installed (#35). If it’s not linked globally, you can set FFMPEG_PATH.

Local Setup

git clone https://github.com/gr2m/voice-of-interconnect.git
cd hoodie
npm install

Start hoodie and pass a valid CouchDB URL with credentials of an admin user

npm start -- --dbUrl=http://admin:[email protected]:5984/

In a 2nd terminal tab, start the app or the visualization. You can also run both at the same time, they use different ports

# start app
npm run app
# start visualization
npm run visualization

If you want to test the build version of the app, run npm run build and open http://localhost:8080 (served directly from Hoodie). The visualization is at http://localhost:8080/visualization/

Connecting to Watson Services

When deployed to Bluemix, the credentials for Speech to Text and AlchemyLanguage are read out directly from VCAP_SERVICES.

To connect to the watson services from other environments, you have to set the following 3 environment variables

1. SPEECH_TO_TEXT_USERNAME
2. SPEECH_TO_TEXT_PASSWORD
3. NATURAL_LANGUAGE_UNDERSTANDING_USERNAME
4. NATURAL_LANGUAGE_UNDERSTANDING_PASSWORD

If they are not set, the services are simulated. There is a default timeout of 3000ms for both the transcription and sentiment analysis. You can change the timeout with WATSON_MOCK_TIMEOUT (ms)

Deployment

See Deployment docs as well as Deploy to Bluemix

Note: the app is redirecting form http:// to https:// unless the host is localhost or 127.0.0.1. The redirect can be disabled with an environment variable: NO_REDIRECT=1

Scripts

For testing integration with Watson Services

# Speech to Text (uploads hardcoded file from public/assets)
SPEECH_TO_TEXT_USERNAME=... SPEECH_TO_TEXT_PASSWORD=... node bin/speech-to-text.js test/fixtures/speech.webm
# .webm and .ogg files are supported
# set FFMPEG_PATH for .webm if ffmpeg binary is not globally linked

# Sentiment (pass any text you like)
NATURAL_LANGUAGE_UNDERSTANDING_USERNAME=... NATURAL_LANGUAGE_UNDERSTANDING_PASSWORD=... node bin/text-to-sentiment.js Marvin is a sore loser

License

Copyright 2017 Neighbourhoodie Software GmbH and Make&Model Inc

Apache 2.0