AlexaOps
Controlling AWS using Alexa
Created by Chris Lennon using reveal.js
Source code available for both control code and the presentation
Introduction
This presentation is to guide you through the creation of an application(skill) for the Alexa platform.
This presentation will give you an overview of (1) the Alexa platform, (2) the usage of a skill to control AWS as an example usecase.
An Overview of Alexa
This section is designed to give an overview of the Alexa ecosystem.
A non technical overview of components.
Alexa, Say Hello.
- Voice and interaction model service by Amazon launched at the end of 2014.
- Hardware devices available from Amazon & third party integrations.
- Application operate around a skill/app model.
Example of skills
Builtin
- Alexa, set a timer for 5 minutes.
- Alexa, add bread to my shopping list.
- Alexa, what is the weather tomorrow?
Skill store
- Alexa, play music on Spotify.
- Alexa, turn off the bedroom lights.
- Alexa, ask Gordon Ramsay what he thinks of my food.
- Alexa, how do I craft a torch using Minecraft helper?
Build your own...
↑ Back To Top
Overview of a skill
This section gives an overview of the components in a skill.
Flow of a Skill
- Stating an action
- Calculating the intent
- Trigger function
- Do action / Get data
- Respond to Alexa service
Setup a skill
This section walks through the creation of a skill.
Requirements
The example skill has the following requirements.
- Alexa Developer account
- AWS account and command line access
- AlexaOps Example Code
- Source on GitHub
- Functions written in Python
- Uses serverless framework (NodeJS) for deployment
AWS CLI
We need to install the AWS CLI and configure it.
aws iam get-user
{
"User": {
...
}
}
Create a new skill
Naming things is hard...
- Invocation name
- Alexa, ask Cloud Control for development status
- Alexa, What is the status of development using Cloud Control
- Language - Setup each locale independently
- Skill ID - amzn1.ask.skill.98a0d5f4-56e6-42bb-9168-63d4235dee83
- Alexa Docs - the good and the bad
Quickstart
Update ALEXA_SKILL_ID variable in serverless.yml
If using the example code you are good to go.
serverless deploy
But what did we actually deploy?...
↑ Back To Top
Interacting with Alexa
Request from Alexa
A JSON payload following a particular format
{
...
"request": {
"type": "IntentRequest",
"intent": {
"name": "GetServices",
"slots": {
"Environment": {
"name": "Environment",
"value": "development"
}
}
}
}
...
}
TODO - Something useful
... this bit is kind of up to you, example in a minute...
Responding to Alexa
Actual speech is written in Speech Synthesis Markup Language (SSML)
I want to tell you a secret.
I am not a real human. .
Can you believe it?
Responding to Alexa ... pt2
A JSON payload following a particular format
{
...
"response": {
"outputSpeech": {
"type": "PlainText",
"text": "Plain text string to speak",
"ssml": "SSML text string to speak "
},
"card": {
"type": "Standard",
"title": "Title of the card",
"content": "Content of a simple card",
"text": "Text content for a standard card",
"image": {
"smallImageUrl": "https://url-to-small-card-image...",
"largeImageUrl": "https://url-to-large-card-image..."
}
},
"reprompt": {
"outputSpeech": {
"type": "PlainText",
"text": "Plain text string to speak",
"ssml": "SSML text string to speak "
}
}
}
...
}
But what about the middle part?
... What can I make it do?
Whatever you want!
My example ties uses Lambda to control AWS
↑ Back To Top
AlexaOps
Running the example
Creating: 'Alexa, list development servers'
Create First Intent
- Open skill in the skill builder
- Create a skill - mentioned earlier
Create First Intent... pt2
Create a slot type
- Create new entry in 'Slot Types'
- Call it Environments
- Add expected values: development, production
Create First Intent... pt3
Setup the intent
- Create a new intent called GetServices
- Create a new intent slot called env of type Environments
Create First Intent... pt4
Populate example phrases
- How many {env} servers are running
- What is the status of the {env} environment
Create Lambda Function
Create a new blank function
- Name it AlexaAWSControl
- Pick a runtime - For this example python
- Create a new role from the Basic Edge Lambda permissions template
- Click 'Create Function'
- Add a trigger of type Alexa Skills Kit
The above is what serverless.yaml handles for us.
Link Lambda and Alexa
Two points to connect
- In the Alexa skill config add the Lambda ARN:
- arn:aws:lambda:eu-west-1:6766234252:function:AlexaAWSControl
- In Lambda add the ALEXA_SKILL_ID environment variable
- amzn1.ask.skill.97a0d6f3-48f6-54cb-3049-64e46e5ffe81
Pick up the request
Basic structure of function
def handler(event, context):
if (event["session"]["application"]["applicationId"] !=
os.environ.get('ALEXA_SKILL_ID')):
raise ValueError("Invalid Application ID")
if event["request"]["type"] == "LaunchRequest":
...
if event["request"]["type"] == "IntentRequest":
...
Route the request
Assuming we are handling an IntentRequest
if event["request"]["intent"]["name"] == "GetServices"
return get_service_status(event["request"]["intent"]["slots"])
Pass the slot data to a function basaed on the intent name.
Action the request
Query AWS using boto3 (built into Lambda)
ec2_resource = boto3.resource('ec2')
def get_service_status(slots):
filters = [{
'Name': name,
'Values': [slots['environment']['name']]
}]
instances = ec2_resource.instances.filter(Filters=filters)
for instance in instances:
if instance.state['Code'] == 16:
online_instances.append(instance.id)
else:
offline_instances.append(instance.id)
A word on permissions
The previous function would need the AmazonEC2ReadOnlyAccess policy to be successful.
- Policies can be added to the role we created for this function
- Please dont just give it root...
Build a response
The goal is to respond with a SSML string.
{
...
"ssml" : "There are 2 development servers online. "
...
}
Considerations to grammar:
- number of server(s)
- they/it is/are online
Testing
Many places to test!
- SSML Speech Simulator - In the 'Test' of Skill Builder
- Function logs - Your app or CloudWatch for Lambda
- Alexa app or webapp
- Serverless local test and Lambda online test
serverless invoke local --function functionName -p test/data.json
↑ Back To Top