Alexa Open Data Toronto

Introduction

Alexa voice interface for Toronto Open Data. Ask "Hey Toronto" something and it will display on a paired web page acting as a Kiosk.

Here's the demo video on YouTube.

Ask Hey Toronto…

• where are Parks

• where are Parks near Yonge and Eglinton

• where are Parks near CN Tower

• where are Skating Rinks

• where are Malls

• where are Swimming Pools

• where are Parking Lots

• where are Parking Garages near me

• where are Parking Garages nearby

• where is the CN Tower

• where is the Eaton Centre

When you ask this question, the Kiosk page will update to show the results, something like this:

This project comes complete with source code for both the Alexa Skill Server, and the Web Server, and instructions how to set up everything.

It should be straight forward to adopt this project to other data sets and / or cities. Note however the project is rather sprawling, as there's a lot of moving parts involved in making this work.

On the Web

This project is available under the name "Hey, Toronto" on Alexa.

You can run the Kiosk Web Page at https://hey.homestar.io. Remember you need both of these together.

Technology Used

Alexa

• Voice Service

• Account Linking

Firebase

• Firebase Database

• Firebase Authentication

Web Development

• React

• Redux

• NodeJS

• Promises

• Google Maps

Server

• Apache HTTP

• Let's Encrypt

Data

• City of Toronto Open Data

• Schema.org

Open Data

This project is built to showcase Toronto Open Data datasets, available here.

However, the techniques we use should be generally transportable to other cities and datasets. We have code for dealing with data in the following formats.

• XML

• JSON

• KML

• Shapefile

The code is easily adapted to new projects.

As an output format, we basically have made a YAML data format using the standard terms defined by another Open Project, Schema.org.

Data Sets

All the data importers are in ./src. Right now, many of them require a separate shell script to download data, but we hope to integrate that into the Node.JS in the future (basically: download a zip file, extract a specific file, convert it to some more tractable format).

• bicycle-stations - BIXI bike stations

• cultural-hotspots - arts an architecture

• green-p-parking - parking

• intersections - major intersections. We read combinations both ways, so Yonge and Eglinton and Eglinton and Yonge will both be available

• libraries - libaries

• parks - parks, skating rinks, etc.

• places-of-interest - tourism-reated

• places-of-worship - churches, mosques, etc.

These data sets are what let us to queries like Where are Skating Rinks near Yonge and Eglinton.

Installation

Environment

Our environment is set up as

• Node JS

• Linux (Centos)

• Apache HTTP

The first is non-negotiable, the other two you can probably subtitute to taste.

Configuration

Do

cp ./config.json.template ./config.json

Fill in all the values.

Libraries

This is needed to parse ESRI shapefile data. It is often used by compile.sh scripts

npm install -g shapefile

This is what we use to run the applications (and to keep them up) in production)

npm install -g pm2

Server Details

Basic Setup

git clone https://github.com/dpjanes/alexa-opendata
npm install
cd ./web
npm install
npm run build

Open Data

Note that we ship with all the compiled data, so you don't need to do this!

First, go to ./src and run the compile.sh file in all subfolders (that have them).

Compile data into our standard format

node ./bin/compile-dst

Make Alexa Skill data

node ./bin/create-slots

Run Servers

pm2 start ./bin/server-alexa.js
pm2 start ./bin/server-web.js

Apache

See the README in ./apache/README.md about how to set up Apache. If you're not using Apache, you can still probably use this as a guide.

Firebase Setup

Create New Project

• go to the Firebase Console

• click on Create New Project

• enter a name for the project, select a payment plan, click on Create Project

Get Settings

This allows our code to access firebase (mostly).

• select your project (if not already selected)

• click on Add Firebase to your web app

• copy the JSON object to config.json in the home folder of this project, specifically in the firebase section. It will look something like this

{ … "firebase": 
{ "apiKey": "AIXXXXXXXXXXXXXXXXXX", 
"authDomain": "alexa-opendata.firebaseapp.com", 
"databaseURL": "https://alexa-opendata.firebaseio.com/", 
"storageBucket": "alexa-opendata-uiueieui.appspot.com" }, 
… }

Set up Authentication Methods

This allows users to log in.

• click on Authentication (left hand side)

• click on Set up Sign-in Method

• enable Google sign in. If you want to add other sign in methods you'll have to look at web/src/actions/auth

• set up the services you want, e.g. Twitter, Facebook, Google…. You'll have to consult the documentation for each of those services to get API keys. Except for Google, which you can do with a click

Add Ouath Redirect Domain

Firebase will only allow named hosts to authenticate. In our case, this is hey.homestar.io.

• click on Authentication (left hand side)

• click on Set up Sign-in Method

• click on Add Domain (further down the page)

• add your domain name.

Firebase-admin

This allows our code to verify that a token coming from Firebase is actually from Firebase. We use this to exchange a short lived token for a longer-lived on that we use to identify users for Alexa.

• follow these instructions

• save the downloaded file to firebase-admin.json in the root folder of this project

Database Rules

• click on Database (left hand side)

• click on Rules

• add the following rules

{ "rules": 
{ "stations": 
{ ".read": true, "$station" : 
{ ".write" : "$station === auth.uid", ".read" : true 
} } } }

These rules allow authenticated users to modify data only under /stations/thier-uid

Alexa

This is a fairly standard Alexa Voice Skill project, of which there's quite a bit of instruction for on the web.

We'll just provide the main details here:

• Create a New (Custom) Skill. It does not use the Audio Player

• In the Interaction Model tab: From ./skill folder, copy the SampleUtterances.txt, the What and Where slots, and the IntentSchema.json

• In the Configuration Tab, select HTTPS and add your URL. In our case, this is https://alexa-opendata.homestar.io/request

• In the Configuration Tab, turn on Account Linking. Add an authorize URL (ours is https://hey.homestar.io/authorize ), and make a client-id (ours in alexa-openid). Copy the client-id and one of the redirect urls to ./config.json

• In the SSL Certificate field, assuming you used Let's Encrypt, select My development endpoint has a certificate from a trusted certificate authority.

If all your servers are up and running, you should be ready to test