Use the test framework
Tock provides extensions to test the bot unitarily.
To use them, it is necessary to add the bot-test library to your project.
With Maven:
<dependency>
<groupId>ai.tock</groupId>
<artifactId>bot-test</artifactId>
<version>24.9.7</version>
<scope>test</scope>
</dependency>
Or gradle :
testCompile 'ai.tock:bot-test:24.9.7'
This entire framework is documented in KDoc format here.
Write a simple test
All of the following examples use JUnit5.
A dedicated extension for Tock and JUnit5 is available.
@RegisterExtension
@JvmField
val ext = TockJUnit5Extension(bot)
@Test
fun `greetings story displays welcome message WHEN locale is fr`() {
ext.send(locale = Locale.FRENCH) {
firstAnswer.assertText("Bienvenue chez le Bot Open Data Sncf! :)")
secondAnswer.assertText("Il s'agit d'un bot de démonstration du framework Tock : https://github.com/theopenconversationkit/tock")
}
}
Since the default connector is Messenger's, it is possible to test the Messenger-specific message in the same way:
lastAnswer.assertMessage(
buttonsTemplate(
It is deliberately very limited, but ask him for directions or departures from a station and see the result!
":)",
postbackButton("Itinéraires", search),
postbackButton("Départs", Departures),
postbackButton("Arrivées", Arrivals)
)
)
To test the message specific to Google Assistant (or any other connector), it is necessary to specify the connector that you want to test:
ext.send(connectorType = gaConnectorType, locale = Locale.FRENCH) {
firstAnswer.assertText(" Bienvenue sur le Bot Open Data Sncf !:)")
secondAnswer.assertText(" Ceci est un bot de démonstration du framework Tock: https://github.com/theopenconversationkit/tock")
lastAnswer.assertMessage(
gaMessage(
"C'est volontairement très limité, mais demandez-lui un itinéraire ou des départs d'une gare et voyez le résultat ! :)",
"Routes",
"Départs",
"Arrivées"
)
)
}
Test a specific Story
In the previous examples, it was not necessary to indicate the story to test (greetings being the default story). Let's say we want the search story, we need to specify the story to test like this:
@Test
fun `search story asks for destination WHEN there is no destination in context`() {
ext.send(intent = search, locale = Locale.FRENCH) {
firstAnswer.assertText("Pöur quelle destination")
}
}
Testing a dialog
It is possible to simulate a complete dialog. For example, we simulate here that the user indicates the destination, then the origin:
@Test
fun `search story asks for origin WHEN there is a destination but no origin in context`() {
ext.send("Je voudrais rechercher un itinéraire", search, locale = Locale.FRENCH) {
firstAnswer.assertText("Pour quelle destination?")
}
ext.send("Lille", indicate_location, locationEntity setTo lille) {
firstBusAnswer.assertText("Pour quelle origine?")
}
ext.send("Paris", indicate_location, locationEntity setTo paris) {
firstBusAnswer.assertText("Quand souhaitez-vous partir?")
}
}
The text in the first parameter of the send method is simply indicative, to help understand the tests. The following parameters are used to define how the NLP will analyze the sentence. For example:
private val lille = PlaceValue(
SncfPlace(
"stop_area",
90,
"Lille Europe",
"Lille Europe (Lille)",
"stop_area:OCE:SA:87223263",
Coordinates(50.638861, 3.075774)
)
)
ext.send("Lille", indicate_location, locationEntity setTo lille)
allows to indicate that the sentence "Lille" is categorized as an intention indicate_location and with a value for the entity location which will be the location lille
Finally it is possible to modify all the values of the mocked bus at initialization. In the following example, we simulate the secondary intention indicate_location in order to indicate the origin:
@Test
fun `search story asks for departure date WHEN there is a destination and an origin but no departure date in context`() {
ext.newRequest("Recherche", search, locale = Locale.FRENCH) {
destination = lille
origin = paris
run()
firstAnswer.assertText("Quand souhaitez-vous partir?")
}
}
Chat with Tock