Closed westlakem closed 5 years ago
Hi Michael, I apologize for the slow response. This is a tricky issue to fix, and I believe represents a significant oversight on our part regarding the dialog behavior.
I do not have an exact time yet for a fix, but in the meantime, have you looked at our end-to-end testing? One advantage of it is that it interacts with the real Alexa. You can see an example of a dialog testing scenario here: https://github.com/alexa/skill-sample-nodejs-college-finder/blob/master/test/e2e/FullSearchHandler.e2e.yml#L26
Our documentation on end-to-end testing is here: https://read.bespoken.io/end-to-end/getting-started
The end-to-end testing does require the skill and interaction model be deployed to test, so it is not without tradeoffs. But if you want to tests a delegated dialog in which you are "taking over" the dialog, it is the best solution we can offer at the moment.
Separately, we will update this ticket once we get this issue scheduled and have a time-frame in mind for a fix.
Looking forward to this fix @jkelvie. As always, thank you for the great work your team is doing! :)
Any updates on this? @dgreene1 @jkelvie
We have not had time to look at this. I recommend using the end-to-end testing, as mentioned above. Alternatively, you can use our new SMAPI Simulation support - that will also be useful for testing skills using the Dialog interface. More information here: https://github.com/bespoken-samples/GuessThePrice/tree/master/test/simulation
Hi @westlakem , our latest version 0.7.1 addresses this issue.
We have scaled back the scope of what our emulator does for dialog management - instead of seeking to emulate the actual interactions from a user's perspective, it focuses on creating the correct payloads that will go back and forth to your skill.
For an example of how to use it, look here: https://github.com/bespoken-samples/skill-sample-nodejs-petmatch/blob/master/lambda/custom/index-test.js
Let me know if you have any questions, and good luck with your testing.
Dialog Flow does not delegate to code between filling slots.
VIRTUAL ALEXA -
SERVER OUTPUT - VIRTUAL ALEXA
The log shows 2 inputs, one where
dialogState == null
and one wheredialogState == 'COMPLETED'
without any state in between:When I test this through the alexa developer portal, I get multiple states in between and the dialog starting state is
STARTED
not null: