Avoid Common Errors in Your Skill Model for Alexa Conversations
To avoid common build errors, review the following list before you build your skill model.
In the Alexa Conversations > Dialogs section of the developer console, make sure that:
- There is a green check mark for every turn in every dialog.
- At least one dialog has at least one of the following:
- A response with an argument
- An API with a required argument and a request template
- An API with an optional argument and an opening sentence informing that argument
- For every required API argument, there is a user turn (and a corresponding utterance set) that asks the user for the argument. For example, if your
GetWeatherAPI requires a city and a date, you must have a user turn (and utterance set) for "What date?" and a separate user turn (and utterance set) for "What city?"
- Each API has at least one dialog that uses the Request Args dialog act. You can't simply have a dialog where the user provides your arguments as part of the initial utterance.
- Within each dialog, the dialog acts follow one of the supported flows described in Work with Dialog Acts in Alexa Conversations.
In the Interaction Model > Intents section of the developer console, make sure that:
- Your interaction model has at least one custom intent with at least one sample utterance. Otherwise, although the skill model might build, the skill session might end prematurely during runtime (for example, with the <Audio only response> message in the Alexa simulator).
- You add the built-in intent
AMAZON.FallbackIntentto the intent list.
In the Alexa Conversations > API Definitions section of the developer console, make sure that each API definition specifies a return type in the Return section at the bottom of the API definition page.
In the Alexa Conversations > Utterance Sets section of the developer console, make sure that your utterance sets don't have empty slot definitions.