Bespoken LLM Benchmark: Does ChatGPT know more than Google and Amazon?
Learn more
November 10, 2017 in Blog

Developer Diary – Building A Cross-Platform Voice App With Jovo

TL;DR Looking for a way to build Alexa skills and Google Assistant actions simultaneously? See how to use Jovo and Bespoken to make a cross platform voice app.

I am very excited to be the newest member of the Bespoken team.

My first assignment – to build a voice app, of course! I’m an experienced full-stack developer, but this is my first voice-first project. And as part of this, I got interested in using the Jovo Framework, which I learned about from my colleagues.

Jovo allows for building cross-platform voice apps for Alexa and Google Assistant – we have been doing a lot of work to integrate our software with it (read more here). And so I was excited to try it out and learn more about the framework as well as Alexa skills and Google Actions. It seemed like a great way to come up to speed on both platforms simultaneously!


First, I had to figure out what I was going to build. When I was a kid I was a big fan of Pokémon, so I thought it would be cool to do something related to that. I decided to create a voice app that requests information on Pokémon from the “Pokédex” (the catalog of Pokémon – there are over 800 now! When did this happen?). I thought it would be a fun first voice experience.

I integrated with a convenient public API of Pokémon to do this, which also has a handy JS/Node client library.

Getting Started With Jovo To Build a Cross-Platform Voice App

Incorporating Jovo into my project was easy – it was as simple as:
npm install jovo-framework-nodejs --save

I also installed the jovo-cli globally, which provides nice helper commands from the console:
npm install jovo-cli -g

Then I followed this helpful guide:

Jovo leverages states and state handlers, an intuitive way to work with skills.

Here is where I got started, with creating a simple Launch Intent handler:

const repromptMessage = 'Ask me which number of pokemon you want to know about';
const errorMessage = 'There was an error with your request, please try again';
const goodbyeMessage = 'Thank you for using my pokedex! Remember to catch them all!';
const pikachuAudio = '';
const handlers = {
'LAUNCH': function () {
var welcome = 'Welcome to my pokedex. ' + repromptMessage;
app.ask(welcome, repromptMessage);
view raw SimpleJovoReply hosted with ❤ by GitHub

When the user requests a Pokédex, I switch their state to “Description State”:

.showImageCard(, description, response.sprites.front_default)
.ask(speech, reprompt);

And to provide back info from the Pokédex, that is managed like so:

'DescriptionState': {
'YesIntent': function () {
var P = new Pokedex();
var number = app.getSessionAttribute('pokemonNo');
var img = app.getSessionAttribute('pokemonImg');
P.getPokemonSpeciesByName(number) // with Promise
.then(function (response) {
var dexEntry = response.flavor_text_entries.filter(function (entry) {
return ( === 'en');
var description = dexEntry[getRandomInt(0, dexEntry.length - 1)].flavor_text;
app.showImageCard(, description, img).tell( + ': ' + description + '. ' + goodbyeMessage);
.catch(function (error) {
'NoIntent': function () {

The code needs little explanation, I hope – all the method signatures from Jovo are self-explanatory and intuitive. In the routine above, we are calling the Pokédex API, getting back info for the selected Pokémon, and then formatting an Alexa reply with a text and a card. Easy!

Testing My Skill for Alexa

To test, I used the new Jovo/Bespoken integration.

By calling: jovo run --bst-proxy, I got this output:

I added the URL displayed to the configuration page in the Alexa developer console (step-by-step walkthrough here).

With that set, I was able to debug and test my app right on my machine, sending requests from Alexa and Google Assistant right to my laptop. I went to the Alexa service simulator to try it out with this phrase: what is the pokemon at 10

I saw this response come back:

Very cool, right?

Getting Setup As A Google Action

Now that I had a working Alexa skill, I turned my attention to Google. Lucky for me, Jovo had a helpful guide for using my voice app there as well – and with the same codebase, no less!

Following these instructions, I was able to setup my app with DialogFlow (previously known as API.AI).

Once everything was configured, I tested it out in the Actions on Google Simulator:

Looks great!

Adding Unit Tests

Now that my app was running well, I decided to add some unit tests. For this, I used our Virtual Alexa project. It’s a very helpful tool that allows for the emulation of the real Alexa behavior.

The unit tests ensure the app handles launch intents and returning Pokédex values correctly – here is a sample:

test("Launches and asks for voltorb, then cancels ", (done) => {
alexa.launch().then((payload) => {
expect(payload.response.outputSpeech.ssml).toContain("Ask me which number of pokemon you want to know about");
return alexa.utter("100");
}).then((payload) => {
return alexa.utter("no");
}).then((payload) => {
expect(payload.response.outputSpeech.ssml).toContain("Thank you");

With this, I can make changes to my Alexa skill with the confidence that I am not going to break anything.

I also added CI and Code Coverage – all part of building an “industrial-strength” voice app. Does that seem excessive for Pokémon? Well, they may seem silly, but they are serious business :-). You can take a look at all of it here, including the source code and app itself.

Wrap Up

Overall, it was a great experience using the Jovo Framework. I was able to easily build a skill and action simultaneously, with a clean and simple API. And now that I know how to develop for both Google Assistant and Amazon Alexa – I look forward to learning more about both, and helping developers like myself build cool voice apps for them.

One response to “Developer Diary – Building A Cross-Platform Voice App With Jovo”

Leave a Reply

Your email address will not be published. Required fields are marked *