10k view:
voice -> parse (via API.AI) -> fullfillment (via API.AI or Lambda)
Detail view:
Voice -> API.AI -> Intents -> Entity (similar to enum) + Action (ex, 'list.tasks') -> Response: in API.AI or webhook (-> Lambda)
For fullfillment, you can use local machine with ngrok or, you can deploy Lambda directly to staging.
Componenet:
- https://github.com/actions-on-google/conversation-components-nodejs/blob/master/api-ai/index.js
- https://developers.google.com/actions/reference/nodejs/Carousel
Sample with Lambda:
the doc is outdates, UI is different now