Building Alexa Skills With Grails

By Ryan Vanderwerf, Software Engineer

July 2016


Alexa Architecture

Recently I gave a talk at GR8Conf EU about building applications for the Amazon Alexa platform. I talked about using Groovy with Lambdas to demonstrate Skills (Alexa Apps), as well as using Grails. I’ll give this talk again at GR8Conf US at the end of July (2016) and have even more good stuff to share. On top of that, after working on this for months, I have the material to present an Alexa workshop at GR8Conf US (Thanks to Collin Harrington for helping with the idea). Good Times!

This article outlines the steps for using a plugin I’ve recently developed to help users crank out Alexa apps (skills or speechlets). Amazon has a promotion right now: if you submit a new skill to their library, you can win a free t-shirt! So maybe you can make something cool with this and get one! You don't even need an Amazon device; you can point your browser to to use their emulator.



Let me get started by explaining how these things work and a little about the devices. Starting a couple years ago, Amazon released a home device/speaker thing called the Echo. It’s a household helper device, with very nice multi-directional microphones and a speaker for playing music. At first, I didn’t get one or understand it; it sounded like a useless toy to me. However, peers gobbled them up and started playing around with them. They weren’t building apps for the Echo, but raved about how cool it was and how it helped automate their home and do useful things. Those same folks started saying they loved the Echo so much that they were buying more units and putting them all over their house. Now my attention was had; of course the first thing that popped into my head was ‘how can I get Groovy to work with this thing?’

Things Alexa Can Do


The first examples they give are all about AWS Lambda functions in Node. However they aren’t fantastic on Java because they have a hard limit of 50mb jars you can upload. Surely you can’t fit all of Grails and dependencies in there to do much. Still we gave it a shot.  I followed Benoit Hédiard’s presentation on Groovy with Lambdas from GGX 2015. I sat down with a buddy of mine, Lee Fox (He will be at GR8Conf US), and we hammered out a bunch of ideas we could build with it - in Groovy, of course!

The first app we got working was a simple Twitter app that combined Groovy, Spring Social and Skills to make a skill that would connect to Twitter and search for tweets, as well as give updates to your timeline. You can find the code here. Good! Now we’re getting somewhere. I did a little digging and found they have some servlet support to make a stand alone web service. Now we can get Grails involved and make this less hideous.

Next we took it a level up. We didn’t want the user to have to hard-code their Twitter API keys to run the app, and you can’t publish a skill that pulls down someone else’s Twitter info. So we built a Grails app that used Spring Security, Oath plugin, and Spring Security UI to allow a UI that allows you to register and enter your own Twitter keys for your account. Source for that is here.

After that, it was time to make a plugin to help you make these skills; that was published while I was out at GR8Conf EU (you can find it here). It works similar to the Quartz plugin, in that you have a CLI command to create a working template of a Skill to get you started. I will show you below how to build a skill using the Grails plugin to get you started making your own Skills in a short about of time.

How Alexa Service Works

The devices themselves (which are currently the Echo, Tap, Dot, and FireTV) don’t do a whole lot. They even publish the source to the devices for you to look at if you choose here. The magic resides on the Amazon side that handles all the voice recognition and handling of requests. Basically what happens is the user initiates an action (skill) on the device, and it goes to the Alexa service to figure out what you want to do. When your app is invoked, it calls back to your app via a JSON HTTPS call and initiates a series of Intents to make your app do things. Your app simply waits for the call and takes appropriate action (via JSON) to do what you want. It’s up to you as the developer to make it do something useful.

Alexa Appkit Architecture


The picture above give you an idea of how it works (from Amazon’s site). Let’s go over Intents and Sample Utterances to have a better idea of what’s going on.


Intents are sort of like simplified Intents on Android, if you have ever done Android coding. They signal an intention for a command to run to do something.

Let’s look at an example:

  1. {
  2. "intents": [
  3. {
  4. "intent": "ResponseIntent",
  5. "slots": [
  6. {
  7. "name": "Answer",
  8. "type": "AMAZON.LITERAL"
  9. }
  10. ]
  11. },
  12. {
  13. "intent": "QuestionCountIntent",
  14. "slots": [
  15. {
  16. "name": "Count",
  17. "type": "AMAZON.NUMBER"
  18. }
  19. ]
  20. },
  21. {
  22. "intent": "AMAZON.HelpIntent"
  23. }
  24. ]
  25. }

Here this application can trigger 3 Intents: ResponseIntent, QuestionCountIntent, and HelpIntent. You use JSON to tell it what these are. You can define ‘slots’, which is the expected responses that are comprised of a name and Type. There are pre-defined data types you can use to help Amazon parse what is said (Number, for example, knows the user is going to say a number. Literal is a simple String. Some Intents don’t need anything as they are ‘built in’, like the help intent (no inputs)).

A little bit about slots - they are entirely optional. You can do things like define a list of allowed responses (custom slots) that are valid things for the user to say. In this case, the ResponseIntent is free form, and we sort out what we want to do by parsing the string we get from Amazon (translated via TTS).

Sample Utterances

Let’s look at another example:

ResponseIntent {test|Answer}
ResponseIntent {last player|Answer}
ResponseIntent {test test|Answer}
ResponseIntent {test test test|Answer}
ResponseIntent {test test test test|Answer}
QuestionCountIntent {Count} questions

Here we see the intent each utterances (what the person says) and map it to an Intent. Amazon will generate common words like conjunctions and ignore things on their end, so you don’t have to mess with handling things like ‘and’, ‘the’, ‘or’ etc. Variables are surrounded by {}. You can use the | operator to specify alternate options. If you want to say something to your skill, it must match the pattern here. For this case, we want to parse multi-world answers to a question. Basically here we support one, two, three or four word responses - anything else will be cut off/ignored. The last item allows the user to answer how many questions they want to be asked.


Let’s Make a Grails App Using the Plugin, and I will Explain More as We Go

The plugin is for Grails 3.x only. Let’s create a new app first:

grails create-app skillsTest

Now lets open build.gradle and add the plugin into the dependencies {} closure:

compile "org.grails.plugins:alexa-skills:0.1.1"

Now let’s create a skill from the command line:

grails create-speechlet SkillsTest
| Rendered template Speechlet.groovy to destination grails-app/speechlets/skillste

Now let’s see what it’s created in grails-app/speechlets:

  1. @Slf4j
  2. class SkillsTestSpeechlet implements GrailsConfigurationAware, Speechlet {
  4. def grailsApplication
  6. Config grailsConfig
  7. def speechletService
  10. def index() {
  11. speechletService.doSpeechlet(request,response, this)
  12. }
  14. /**
  15.   * This is called when the session is started
  16.   * Add an initialization setup for the session here
  17.   * @param request SessionStartedRequest
  18.   * @param session Session
  19.   * @throws SpeechletException
  20.   */
  21. public void onSessionStarted(final SessionStartedRequest request, final Session session)
  22. throws SpeechletException {
  23."onSessionStarted requestId={}, sessionId={}", request.getRequestId(),
  24. session.getSessionId())
  26. }
  28. /**
  29.   * This is called when the skill/speechlet is launched on Alexa
  30.   * @param request LaunchRequest
  31.   * @param session Session
  32.   * @return
  33.   * @throws SpeechletException
  34.   */
  35. public SpeechletResponse onLaunch(final LaunchRequest request, final Session session)
  36. throws SpeechletException {
  37."onLaunch requestId={}, sessionId={}", request.getRequestId(),
  38. session.getSessionId())
  40. return getWelcomeResponse()
  41. }
  43. /**
  44.   * This is the method fired when an intent is called
  45.   *
  46.   * @param request IntentRequest intent called from Alexa
  47.   * @param session Session
  48.   * @return SpeechletResponse tell or ask type
  49.   * @throws SpeechletException
  50.   */
  51. public SpeechletResponse onIntent(final IntentRequest request, final Session session)
  52. throws SpeechletException {
  53."onIntent requestId={}, sessionId={}", request.getRequestId(),
  54. session.getSessionId())
  56. log.debug("invoking intent:${intentName}")
  57. PlainTextOutputSpeech speech = new PlainTextOutputSpeech()
  58. // Create the Simple card content.
  59. SimpleCard card = new SimpleCard(title:"Twitter Search Results")
  60. def speechText = "I will say something"
  61. def cardText = "I will print something"
  62. // Create the plain text output.
  63. speech.setText(speechText)
  64. card.setContent(cardText)
  65. SpeechletResponse.newTellResponse(speech, card)
  67. }
  68. /**
  69.   * Grails config is injected here for configuration of your speechlet
  70.   * @param co Config
  71.   */
  72. void setConfiguration(Config co) {
  73. this.grailsConfig = co
  74. }
  76. /**
  77.   * this is where you do session cleanup
  78.   * @param request SessionEndedRequest
  79.   * @param session
  80.   * @throws SpeechletException
  81.   */
  82. public void onSessionEnded(final SessionEndedRequest request, final Session session)
  83. throws SpeechletException {
  84."onSessionEnded requestId={}, sessionId={}", request.getRequestId(),
  85. session.getSessionId())
  86. // any cleanup logic goes here
  87. }
  89. SpeechletResponse getWelcomeResponse() {
  90. String speechText = "Say something when the skill starts"
  92. // Create the Simple card content.
  93. SimpleCard card = new SimpleCard(title: "YourWelcomeCardTitle", content: speechText)
  95. // Create the plain text output.
  96. PlainTextOutputSpeech speech = new PlainTextOutputSpeech(text:speechText)
  98. // Create reprompt
  99. Reprompt reprompt = new Reprompt(outputSpeech: speech)
  101. SpeechletResponse.newAskResponse(speech, reprompt, card)
  102. }
  104. /**
  105.   * default responder when a help intent is launched on how to use your speechlet
  106.   * @return
  107.   */
  108. SpeechletResponse getHelpResponse() {
  109. String speechText = "Say something when the skill need help"
  110. // Create the Simple card content.
  111. SimpleCard card = new SimpleCard(title:"YourHelpCardTitle",
  112. content:speechText)
  113. // Create the plain text output.
  114. PlainTextOutputSpeech speech = new PlainTextOutputSpeech(text:speechText)
  115. // Create reprompt
  116. Reprompt reprompt = new Reprompt(outputSpeech: speech)
  117. SpeechletResponse.newAskResponse(speech, reprompt, card)
  118. }
  120. /**
  121.   * if you are using account linking, this is used to send a card with a link to your app to get started
  122.   * @param session
  123.   * @return
  124.   */
  125. SpeechletResponse createLinkCard(Session session) {
  127. String speechText = "Please use the alexa app to link account."
  128. // Create the Simple card content.
  129. LinkAccountCard card = new LinkAccountCard()
  130. // Create the plain text output.
  131. PlainTextOutputSpeech speech = new PlainTextOutputSpeech(text:speechText)
  132. log.debug("Session ID=${session.sessionId}")
  133. // Create reprompt
  134. Reprompt reprompt = new Reprompt(outputSpeech: speech)
  135. SpeechletResponse.newTellResponse(speech, card)
  136. }
  138. }

Also the plugin will generate a Controller class embedded in your Speechlet file. If you are using Spring Security, you will want to make sure that uri is accessible to the outside so the Alexa service can contact it (there are some requirements I’ll fill you in on later):

  1. /**
  2.  * this controller handles incoming requests - be sure to white list it with SpringSecurity
  3.  * or whatever you are using
  4.  */
  5. @Controller
  6. class SkillsTestController {
  8. def speechletService
  9. def skillsTestSpeechlet
  12. def index() {
  13. speechletService.doSpeechlet(request,response, skillsTestSpeechlet)
  14. }
  16. }

The speechlet artefact will be registered as a Spring bean so it’s automatically injected. There is also a service the plugin provides to handle the boring stuff, like verifying the request, checking the app ID (more on that later) and the plumbing that calls your skill.

Built-in Events

Looking at the code example above; you can see it made several methods for you. These are part of the skill (speechlet) lifecycle. The first one is:


This allows you to store variables for the duration of the session (the interactions as a whole of the app for that time). You can do some setup here and store variables you can use later. This is technically optional for you to implement.


This is called when you invoke the skill. When you say ‘Alexa open skillTest’ etc., this is your chance to say an opening message about your app, what it does, or what they will need to do.


This is the meat and is required to be implemented. When your sample utterances map to an Intent when the user says something, this is invoked. Here you should generate a card that will appear in the Alexa app on your phone (you can also get to this on your local network via browser by going to ‘’). Cards are similar to Android cards (more popular in Android Wear) that simply show a message to the user so they can see what is going on. You can make several kinds of cards, which are Simple, Standard, and LinkAccount. Here we have a switch statement to figure out what Intent to process and call the code for the appropriate intent.


This is the last one (optional) where you can clean up and session resources you might have created, like database records for the run.

I’ve added a few other helper methods to see how to render a help response, link an account, and get some setup details from the Grails config. As a first pass, just try to get the default template working, then start to change things like changing the text, adding an Intent, etc.

Set Up Your App on Amazon Developer Portal

This is separate from AWS. I am not aware of any APIs that will create all of this for you, so you have to sign up for an account and do this manually for each skill you want to run.

Developer Portal 1


Add Alexa Skill


Developer Portal 3


Developer Portal 4


Developer Portal 5


Now note the application ID it gives you. You will need to add this to the application.groovy/yml file so the application will know the app ID and accept it.

Copy the application ID on the first tab ‘SKILL INFORMATION’, and paste that into application.groovy.

alexaSkills.disableVerificationCheck = true // helpful for debugging or replay a command via curl
alexaSkills.serializeRequests = true // this logs the requests to disk to help you debug
alexaSkills.serializeRequestsOutputPath = "/tmp/"

Build and deploy your war file to your server (btw, it must be port 443 HTTPS, no exceptions).

Test your app

Now try it on your Echo/Alexa device. Say either ‘start’ or ‘open’ and the invocation name you gave the app, and follow the prompts! You can also use the test function on the portal itself.


Debugging can be very frustrating. Make sure to turn your log level to debug. In the Grails config, there is an option called ‘serializeRequests’ and an output path for them. This allows you to capture the request that came from Amazon. If you are trying to test a fix for a bug, you can replay this via CURL. The files will look like this:

-rw-r--r-- 1 tomcat tomcat      598 Jun  6 22:45 speechlet-1465249551692.out
-rw-r--r-- 1 tomcat tomcat      636 Jun  6 22:45 speechlet-1465249557538.out
-rw-r--r-- 1 tomcat tomcat      598 Jun  6 22:46 speechlet-1465249601675.out
-rw-r--r-- 1 tomcat tomcat      636 Jun  6 22:46 speechlet-1465249607216.out

The build-in security provided by the plugin and underlying library will now allow you to reuse a request because the hash and timestamps are too old (to prevent this type of attack called a ‘replay attack’). You can disable this check for dev purposes with the ‘disableVerificationCheck’ Grails config value. Now you can replay a file via CURL back to your server to avoid the whole voice interaction to test that one case (and test it locally!):

curl -vX POST http://localhost:8080/test/index -d @speechlet-14641464378122470.out --header 

If you save enough requests of a normal interaction, you could write some functional tests that replay this as part of a test suite or just be able to dev against them locally a bit.

Advanced Stuff - Full UI with Spring Security, OAuth

If you want a UI for the user and to use account linking, an easy path is to add Spring Security, Spring Security UI, Spring Security OAuth grails plugins. You can see an example of how to do this here.

Advanced Stuff - Say Sample Audio Clips in Your Skill

There is a supported markup called SSML which allows up to 90 seconds of low quality mp3 sound clips to play. Their settings/requirements are quite picky to work:

Example SSML:

  1. <speak>
  2. <audio src="\"\"/"> ${speechText}
  3. </audio>
  4. </speak>


The Alexa service is a great invention that is catching on. Google and Apple are dipping their toes into the market. I see the home-based star trek experience being a reality for everyone in just a few years. Already, my wife (who swore she never would) uses it, and my 4 year old asks it to tell her jokes all the time. I use it to control my lights very often. The sky is the limit, and these tools can be useful in the workplace too. So I encourage you to get out there and build some neat stuff for Alexa with Grails!

Useful links


The Software Engineering Tech Trends is a monthly newsletter featuring emerging trends in software engineering.


© Copyright Object Computing, Inc. 1993, 2016. All rights reserved

WebSanity Top Secret