Skip to content
- Invocation name
- Use cases
- Make a request = “Alexa, ask <invocation name> for blah blah…”
- Use predefined phrases = “Alexa, open<invocation name>…”
- Ask / Open / Play / Resume / Start
- Direct invocation = “Alexa, <invocation name>…”
- CANNOT use launch phrases, article, connecting words
- app, weather, the, a, and…
- Can be changed during development, cannot be changed after a skill is certified and published
- Interaction models
- Pre-built models
- Music
- Video
- Home Appliances
- Baby activities
- Lists
- Custom models
- Voice Action
- Make a request = click a button
- Collect information = show a list
- Provide information = click an element in a list
- Sample utterances = templates for voice input, mapped to intents
- “What’s the temperature now?”
- “temperature, please?”
- Slot = Variables in input, consistent across sample utterances
- “
{Amount}
US dollars in Korean won”
- Extended Slot = Add more values to a slot
- Slot elicitation = Alexa asks a question for a slot value
- Alexa: “Where city are you leaving from?” User:”
{FromCity}
“
- Slot confirmation = Alexa repeats slot value and user responds “Yes / No”
- Multi-turn
- set
shouldEndSession
to false
in response object
- Voice with Display = also display information when Alexa-enabled device has a screen
- Card
- Simple = text only
- Standard = text + image
- LinkAccount = for linking account only
- AskForPermission
- Intents = actions user may take
- Standard intents
- Fallback intents
- Handle unexpected slot value or response
- Help intent
- In-skill Purchase
- One-time
- Consumable
- Subscription
- Alexa for Business
- Skill Endpoint
- Must be SSL and signed by a trusted CA
- Account Linking = SSO for Alexa
- OAuth 2.0
- Grants
- Custom skills can use authorization code or implicit grant
- Music, home appliances must use auth code grant
- Scope = used to avoid collecting sensitive personal info
- Customer Profile API = use information from active default Alexa profile
- SSML = Speech Synthesis Markup Language, add emotion to speech
- prosody = volume, pitch, speed
- audio = play audio
- MP3, 48kbps, 280s, <=24khz, must be publicly accessible with HTTPS
- emphasis
&
should be escaped to &
just like XML
- State management
- Application level = global state
- Session level
- Between blue light on and off = during a skill invocation
- User level
- Alexa Skills Kit = ASK
- AWS Service that extends Alexa skills
- Lambda = most used for Alexa
- Alexa timeout is 8s, Lambda timeout defaults to 3s
- CloudWatch
- Mostly you use CloudWatch logs to collect Lambda logs
- Alarm to trigger alarm when something goes wrong
- S3
- APIs
- IntentRequest / Display.ElementSelected will return an
outputSpeech
element
- Beta test = Green/blue for Alexa
- Hooks
Like this:
Like Loading...