Skip to content

Latest commit

 

History

History
285 lines (175 loc) · 9.7 KB

AWS Certified Alexa Skill Builder - Specialty.md

File metadata and controls

285 lines (175 loc) · 9.7 KB
Raw

AWS Certified Alexa Skill Builder - Specialty

AWS Account

https://aws.amazon.com/free/

Alexa Developer Console

https://developer.amazon.com/alexa/console/ask

Skill Types

Custom Skills - complete control over the UX, define your own custom interaction model

Smart Home Skills - communicate with smart home devices using prebuilt model

Flash Briefing Skills - short overview of news or other content

Video Skills - provide video content

Music Skills - provide audio content

List Skills - for providing CRUD to a list

Design Patterns

Adaptability - adaptable skill understands what a user says appropriately

Multiple Utterances Account for Over-Answering (too much information) Request Additional Information (re-prompt the user) Handle Corrections Plan for Errors (gracefully handle errors, don't use generic responses) Anticipate for Alexa not understanding (feature not available etc.)

Be Adaptable

Personalization - personalized skill remembers interactions and information about the user

Custom welcome message (explain the skill for first time, second time say welcome back etc.) Remember user information Remember user interactions (persist to database)

Be Personal

Availability - available skill guides the user and keeps all options open

Response Time Limits - 8 second time limit for user, prompt after the time limit Effectively Manage lists - simple options, use Speech Synthesis Markup Language (SSML)

Complete Task and End the Session - end session after you fulfill a user's request

Be Available

Relatability - relatable skill allows the user to feel like they are having a conversation

Don't make it formal, vary responses, use transition and timeline markers

Be Relatable

Voice Considerations

Think voice first and screen second (not a requirement), provide context and mind the 8 seconds response rule (no information overload) Leverage Dialog Management Prepare for the global marketplace (localization) Speechcons and SSML

Speechcons are special words and phrases that Alexa pronounces more expressively.

Skill Design

Define Purpose and Goals Who are your users? What can users do with the skill? What information you need?

Identify customer needs and how your skill will address those needs

Steps

Write a script - dialog between Alexa and users Develop a storyboard - add variations and error cases to the script Interaction model - use Alexa Skills Kit (ASK) to implement logic and voice interface (Alexa Developer Console and Command Line Interface can be use to define it)

Custom (complete control) and Pre-built (requests and utterances are defined, less control) interaction models

Script and storyboard document the information flow

Flow consists of:

  • Utterance
  • Situation
  • Response
  • Prompt

Interaction model

consist of 4 components:

Intents - action that fulfills a user request (leverage built in intents to save time) Utterances - spoken phrases the user says Custom Slots - possible values Dialog Model - defines the steps for multi turn conversation

When you say Alexa (wake word) everything after is recorded and streamed to the cloud via WiFi. Alexa Voice Service (AVS) consists of Services, APIs and uses Machine Learning and Natural Language Understanding to process the request.

Frontend - Alexa Developer Console (Alexa Skills Kit (ASK)) Backend - AWS or whatever endpoint you prefer (HTTPS), verify is request it from Alexa, adhere to ASK JSON format

Creation Process

Voice User Interface (VUI)

defines the intents (user requests) and words (utterances)

Fulfillment Logic (backend, typically LAMBDA) Connect VUI to Code Building & Testing (thru console and real device) Distribution (store, business, personal) Certification (Amazon tests and verifies)

Alexa Skills Kit (ASK)

Collection of APIs, Tools, Documentation and Code Samples

Skills Creation Tools

ASK Command Line Interface (ASK CLI)

npm ask-cli

bash npm install -g ask-cli

uses SMAPI behind the scenes

ASK CLI Command Reference

Skill Management API (SMAPI)

Manage your skill programmatically using API root api.amazonalexa.com

Alexa Developer Console

Nice UI for creating skills, instead of using CLI

AWS Tools

AWS CodeStar, AWS CodePipeline, AWS CloudFormation

Skill Flow Builder

GUI for designers and story writers (content) and developers to implement those

Set up Skill Flow Builder as a Developer

npm install --global @alexa-games/sfb-cli
npx alexa-sfb
 npx alexa-sfb vscode
 npx alexa-sfb new my_best_story
 npx alexa-sfb simulate <your_project_path>

Invocation Name

Only used for Custom Skill, not needed for prebuilt model

Invocation Name Requirements

Invoking using:

  • IntentRequest (Alexa - tell APP to do something)
  • LaunchRequest (no intent - Alexa open App)
  • CanFulfillIntentRequest (no name, Alexa searches for a skill to match request)

Intent is an action that fulfills user's request (has a NAME and list of UTTERANCES)

2 types of Intents:

  • Built-in

Standard CancelIntent HelpIntent StopIntent FallbackIntent (no match) NextIntent YesIntent NoIntent PauseIntent SearchAction (lookup information)

Standard with screen NavigateHomeIntent (ends skill session, user leaves the skill)

Standard Built-in Intents

  • Custom (complete control)

Slot(s) are values passed to the utterance

Slot Type Reference

Interfaces

Audio Player - streaming audio and monitoring playback Display/ DisplayTemplate Video App - streaming video files GadgetController / GameEngine CanFulfillIntentRequest - no name Alexa Presentation Language Auto Delegation / Dialog

List of Alexa Interfaces and Supported Languages

Request Types

  • LaunchRequest - opening the skill
  • IntentRequest - sent when intent corresponds with user query
  • SessionEndedRequest - when session ends (any reason)
  • CanFulFillIntentRequest - when Alexa is querying if there is a skill that can fulfill the intent (can opting but then need to implement it for all intents)

Handlers

Request Handlers - handle incoming request and response Exception Handlers - for errors Handler Classes - implement the abstract class AbstractRequestHandler and its two methods, can_handle() and handle()

Leverage userId and deviceId for personalization and segmentation along with DynamoDB

Screen Devices

Available standard built-in intents for Alexa-enabled devices with a screen

Must implement Amazon.PreviousIntent and Amazon.NextIntent

Card types:

  • Simple (title and content)
  • Standard (title, text and image)
  • Linked Account
  • AskForPermissionsConsent

Cards displayed on Alexa-enabled devices with a screen

When to include cards, and how often?

Limit the number of cards Too many will take the user out of the voice experience. Avoid pushing cards with every response, unless it is absolutely necessary.

For screen devices, voice should remain the primary form of interaction, and as much as possible, the user should be able to navigate through the skill strictly by voice.

Display templates:

  • Body Templates (1 - 7) - can't select images
  • List Templates () - can select images
  • Images

Display Template Reference

Display Interface Reference

Display Template is used if both Display Template and Card are in the response

check the {device: supportedInterfaces}

Alexa Presentation Language (APL)

Offer more control than Display Template(s)

Understand Alexa Presentation Language (APL)

Speech Synthesis Markup Language (SSML)

Supported SSML tags

Leverage Voice & Tone tab to experiment with SSML tags

PlaybackController

The Alexa.PlaybackController interface describes messages that are used to play, stop, and navigate playback for audio or video content.