sidecus / zyin.intentbot Goto Github PK

NLP services like Luis supports entity extraction as part of the intent understanding.
It'll be better to connect PromptManager with those entities so that we don't have to ask user the info again.

It'll require IntentContext base class to have a new method which takes a parameter like a Dictionary and being able to automatically assign the entity value to corresponding user input properties.

Not too hard but didn't have time to implement it yet.

...just being too lazy

OAuthPromptSettings.TimeOut configures how long the OAuth prompt will be waiting for user input. We are using 30 seconds right now. If a user doesn't authentication in time, next message will result in an empty TokenResponse.

We have code to save some user info based on the token (OAuthDialog.SaveUserInfoAsync). This code right now doesn't handle null token well so we end up with a NullReferenceException.

We should return a user friendly message instead here.

I am just too lazy for this leisure time toy. If this turns out to be useful for others, I'll think about adding unit tests for the library to ensure quality.

I just could not resist caching the user token, so I added some code to do that in user states.
Yeah it may not be that secure, it's not suggested.
It's just faster for "simple" single turn intents.

Opening this to track giving library user an option to turn the token caching behavior off.

I create some simple classes to help json based adaptive card formatting.
Later I realized that AdaptiveCard team is working on something called AdaptiveCard templates which does exactly the same thing.

It's still in early stage, but we should switch to the new AdaptiveCard templates when it's stable.