Nesting input-groups - bixby

Question

So for one of my actions I need the following combination of input from a user:
A OR (B AND\OR C)
Alternatively:
A OR (B AND C) OR (B OR C)
So the following would be valid combinations of inputs:
A
B
C
B & C
I have tried to implement the following:
action (blah) {
description (blah)
type (blah)
collect {
input-group (inputs) {
requires (OneOf)
collect {
input (a) {
type (A)
max (One)
}
input-group (bAndOrC) {
requires (OneOrMoreOf)
collect {
input (b) {
type (B)
max (One)
}
input (c) {
type (C)
max (One)
}
}
}
}
}
}
output (blah)
}
However this gives me an error "illegal combination of max-one constraint (inputs) and min-required member (bAndOrC)".
This error goes away if I change OneOrMoreOf to ZeroOrMoreOf. However this seems like it would mean the user could provide no inputs and it would be considered valid. Or maybe I am misunderstanding and the OneOf in the outer input-group is "inherited" by the inner input-group such that it is effectively OneOrMoreOf?

Answer 1

The easiest way to do this would be to have 2 actions:
Action1 would accept A as the sole input
Action2 would have an input-group that would require OneOrMoreOf and collect B and C as inputs within the group.
These two actions will correctly address the inputs A alone, B alone, C alone, and B&C.
It is important to train multiple types of utterances in your training file to ensure Bixby has enough context to discern between InputA, InputB, and InputC values.
For example, "Input A is 12" and "A is 12" should both be trained to teach Bixby that the context lies more with "A" rather than anything else.
Action1 model:
action (Action1) {
type (Search)
description (__DESCRIPTION__)
collect {
input (inputA) {
type (InputA)
min (Required) max (One)
}
}
output (GenericOutput)
}
Action2 model:
action (Action2) {
type (Search)
description (__DESCRIPTION__)
collect {
input-group (BAndOrC) {
requires (OneOrMoreOf)
collect {
input (inputB) {
type (InputB)
min (Optional) max (One)
}
input (inputC) {
type (InputC)
min (Optional) max (One)
}
}
}
}
output (GenericOutput)
}

Answer 2

So it turns out that the ZeroOrMoreOf nested within a OneOf effectively turns it into a OneOrMoreOf. In the simulator, if I provided no inputs, it prompted me for an input rather than trying to complete the action with no inputs.
However, another problem I ran into was that, in some cases, Bixby couldn't correctly list valid inputs for the user to choose from if too many were provided. It would do fine if, for example, the input was a1, a2, a3. In this case, it would prompt for the user to select one of the three. It would also do fine if, for example, the input was b1, b2, c1, c2. In this case it would prompt for the user to select for one the the B inputs and one of the C inputs. However if the input was for example a1, b1, c1 it would chose the input to be a1, ignoring b1 and c1 without a prompt. If the input was a1, a2, b1, c1, it would prompt the user to choose between a1 and a2, completely ignoring b1 and c1. To solve this I created a preliminary action that would accept any combination and number of A, B, and C and then output all the possible combination for the user to choose from and then use that as an input to the actual action. So my preliminary action looked like
action (preliminary) {
description (blah)
type (blah)
collect {
input-group (inputs) {
requires (OneOrMoreOf)
collect {
input (a) {
type (A)
max (Many)
}
input (b) {
type (B)
max (Many)
}
input (c) {
type (C)
max (Many)
}
}
}
}
output (PreliminaryOutput)
}
The Javascript for this action would generate a list of possible options for the user to choose from. For example if the input was a1, a2, b1, b2, c1, c2 it would list:
a1
a2
b1 c1
b1 c2
b2 c1
b2 c2
The user would chose one of these which went into the actual action:
action (actual) {
description (blah)
type (blah)
collect {
input (preliminary) {
type (PreliminaryOutput)
min (Required)
max (One)
}
output (RealOutput)
}