A sample Watson Work cognitive app that listens to messages posted to a space in IBM Watson Workspace, understands the natural language conversation happening in the space and posts helpful weather information as needed in the conversation.
The Watson Work platform provides spaces for people to exchange messages in conversations. This sample app shows the following aspects of a Watson Work cognitive application:
- how to implement a Watson Work application and use Watson Work APIs using Node.js;
- how to authenticate and obtain an OAuth token, listen to a conversation in a space, receive messages on a Webhook endpoint, and send messages back to the conversation;
- how to use the Watson Work Services and Watson Conversation cognitive capabilities to understand natural language, identify domain specific user intents, recognize entities such as locations and cities for example, and act based on previously identified intents and entities;
- how to handle a multi-turn conversation and keep track of what's being said across multiple messages.
To try the sample app do the following:
If you want to give the sample app a quick try using [Bluemix] (https://bluemix.net), you can simply get it deployed to Bluemix straight from Github without even having to download it to your local development environment and build it yourself. Just click the button below:
Once that's done, go to your
Bluemix Dashboard. The
app you've just deployed should be listed on that page. Write down its
route public URL (usually https://<bluemix app name>.mybluemix.net)
as you will need it later to register the app's Webhook endpoint with
the Watson Work platform.
You can skip this if you've just deployed the app directly to Bluemix.
To build the app in your local development environment, follow these steps:
Install Node.js 6+.
In a terminal window, do the following:
# For more verbose output
export DEBUG=watsonwork-*
# Get the code
git clone https://github.com/watsonwork/watsonwork-weather
# Build the app
cd watsonwork-weather
npm run buildThe sample Weather app uses Watson Conversation to understand natural language and provide a natural language conversational interface, so you need to configure a Watson Conversation Bluemix service for it.
Go to the Bluemix Watson Dashboard and create a Watson Conversation service.
Note the Watson Conversation service user name and password, as you will need to configure the Weather app with them.
From the Watson Conversation service page click Launch tool to open the Watson Conversation tooling, and import watson.json into a new Watson Conversation workspace.
Note the Watson Conversation workspace id, as you will need to configure the Weather app with it.
The sample Weather app uses the Weather Company Data API to retrieve weather information, so you need to configure a Weather Company Data Bluemix service for it.
Go to the [Bluemix Data & Analytics Dashboard] (https://console.ng.bluemix.net/catalog/?category=data) and create a Weather Company Data service.
Note the Weather Company Data service user name and password, as you will need to configure the Weather app with them.
In your Web browser, go to [Watson Work Services / Apps] (https://workspace.ibm.com/developer/apps) and add a new app named Weather with a Webhook configured for message-created and message-annotation-added events.
Set the Webhook Callback URL to a public URL targeting the server where
you're planning to run the sample app,
https://<your server hostname>/weather for example, or
https://<bluemix app name>.mybluemix.net/weather if you've deployed it
to Bluemix.
Configure the Make It Cognitive section of the app to use your Watson Conversation workspace, user and password.
Save the app and write down its app id, app secret and Webhook secret.
Go to your Bluemix Dashboard, select your app and under Runtime / Environment Variables / User Defined, add the following variables:
WEATHER_APP_ID: <the Weather app id>
WEATHER_APP_SECRET: <the Weather app secret>
WEATHER_WEBHOOK_SECRET: <the Weather Webhook secret>
WEATHER_TWC_USER: <your Weather company service user>
WEATHER_TWC_PASSWORD: <your Weather company service password>
DEBUG: watsonwork-*
Click the > Start button to start the app.
If you've followed the above steps to deploy the app to Bluemix, it is now also set up as a project in the Bluemix DevOps Services Web IDE, allowing you to edit and manage the app directly from within the IDE.
You can skip this step if you're not planning to use that Web IDE. To enable the app to be launched directly from the IDE, edit its Launch Configuration and under Manifest Settings, set its launch Command to:
npm install babel-cli@6.10.1 && npm run babel && npm start
You can skip this if you've just started the app on Bluemix.
In the terminal window, do the following:
# Configure the app id and app secret
export WEATHER_APP_ID=<the Weather app id>
export WEATHER_APP_SECRET=<the Weather app secret>
export WEATHER_WEBHOOK_SECRET=<the Weather Webhook secret>
export WEATHER_TWC_USER: <your Weather company service user>
export WEATHER_TWC_PASSWORD: <your Weather company service password>
The Watson Work platform requires Webhook endpoints to use HTTPS. The sample app listens on HTTPS port 443 and can be configured to use an SSL certificate like follows:
# Configure the SSL certificate
export SSLCERT=<path to your SSL certificate in PEM format>
export SSLKEY=<path to your SSL certificate key in PEM format>
# Start the app
npm start
You can also use a different HTTPS port number and a self-signed certificate, like follows:
# Configure the HTTPS port number
export SSLPORT=8443
# Generate a self-signed SSL certificate with /CN set to your server's
# FQDN (fully qualified domain name), www.yourcompany.com for example
openssl req -nodes -new -x509 -keyout server.key -out server.crt -subj "/CN=your server's FQDN"
export SSLCERT=server.crt
export SSLKEY=server.key
# Start the app
npm start
If you're running behind a HTTPS proxy, you may want to have the app listen on HTTP instead to let the proxy handle the HTTPS to HTTP conversion, like follows:
# Configure the HTTP port
export PORT=8080
# Start the app
npm start
Finally, if the app is running on your development machine and you don't want to set up a public IP and domain name for it yourself, you can also use one the tunnel tools popular for Webhook development like localtunnel or ngrok for example.
Here's how to use a tunnel with localtunnel:
# Install the localtunnel module
npm install -g localtunnel
# Set up a tunnel from https://<subdomain name>.localtunnel.me
# to localhost:8080
lt --subdomain <pick a subdomain name> --port 8080
# Configure the app HTTP port
# No need for HTTPS here as localtunnel handles it
export PORT=8080
# Start the app
npm start
You can now go back to
Watson Work Services / Apps,
edit the Weather app and set its Webhook Callback URL to
https://<subdomain name>.localtunnel.me/weather.
Now that the app is running and listening for HTTPS requests at a public URL, you're ready to enable its Webhook on the Watson Work platform.
Go back to Watson Work Services / Apps, edit the Weather app and set its Webhook to Enabled. Watson Work will ping the app Webhook callback URL with a verification challenge request to check that it's up and responding correctly.
The sample app will respond to that challenge request and output the following log:
watsonwork-weather-app Got Webhook verification challenge
You're now ready to chat with the sample app!
Go to Watson Workspace and create a space named Examples, then open the Apps tab for that space and add the Weather app to it.
In the Examples space, say "Is it raining in San Francisco?".
The Weather app will respond with a message asking you to confirm that you're interested in the weather in San Francisco: "Hey [your name], I think you're looking for the weather conditions in San Francisco. Is that correct?".
Say "yes".
The Weather app will then respond with the weather conditions in San Francisco, like this for example: "San Francisco, CA, 48F Feels like 41F, Fair"
The sample project source tree is organized as follows:
README.md - this README
package.json - Node.js package definition
watson.json - Watson Conversation training configuration
src/ - Javascript sources
app.js - main app conversation handling script
events.js - routes Webhook events to app logic
messages.js - reads and sends messages
graphql.js - runs GraphQL queries
oauth.js - obtains OAuth tokens for the app
sign.js - signs and verifies Webhook requests and responses
state.js - stores conversation state in a database
users.js - queries user profile info
weather.js - gets weather info from The Weather Company
geocode.js - gets the geolocation of a city
ssl.js - configures the app to use an SSL certificate
test/ - unit testsThe app leverages the Watson Work Services and Watson Conversation cognitive capabilities to understand natural language intents, entities, and act based on previous and current entities and intents.
It uses the Watson Work OAuth API to authenticate and get an OAuth token.
It implements a Webhook endpoint according to the Watson Work Webhook API to listen to conversations in a space and receive messages and message annotations.
Finally, it uses the Watson Work Spaces API to send back weather information messages to the space.
Pull requests welcome!