Hi there!
I have built (a quite complex) Jovo Model for Alexa and Google Conversational Actions. Some slot types are different depending which platform is used. When i build the model for Google, the annotations for my types in the sample utterances are not taken from my custom slot.
Here is the (simplified) model adapted from conversational-actions-hello-world:
{
"invocation": "my test app",
"intents": [
{
"name": "MyNameIsIntent",
"phrases": ["{name}", "my name is {name}", "i am {name}", "you can call me {name}"],
"inputs": [
{
"name": "name",
"type": {
"googleAssistant": "NameInputType"
}
}
]
}
],
"googleAssistant": {
"custom": {
"types":[
{
"name": "NameInputType",
"values": [
{
"value": "max"
},
{
"value": "jane"
},
{
"value": "joe"
}
]
}
]
}
}
}
jovo build --clean -p googleAction
is resulting in:
"trainingPhrases":
- "($name 'name' auto=true)"
- "my name is ($name 'name' auto=true)"
- "i am ($name 'name' auto=true)"
- "you can call me ($name 'name' auto=true)"
instead of:
"trainingPhrases":
- "($name 'max' auto=true)"
- "my name is ($name 'max' auto=true)"
- "i am ($name 'max' auto=true)"
- "you can call me ($name 'max' auto=true)"
A further improvement would be if the sample utterances would be filled with random slot values (Google Actions Builder console is giving a warning here at the moment):
"trainingPhrases":
- "($name 'max' auto=true)"
- "my name is ($name 'jane' auto=true)"
- "i am ($name 'joe' auto=true)"
- "you can call me ($name 'max' auto=true)"
Any chance you could fix (and improve) this?
Best,
Dominik.