search

libertymutual / chatpickle

Conversation Tests for Chatbots
Apache License 2.0
10 stars 6 forks source link

Allow user silent response #42

Open dtbell99 opened 3 years ago

dtbell99 commented 3 years ago

Is it possible to get ChatPickle to allow a silent response from the user? Suggestion below passing to response with silence to bot.

bot: what can I help you with? user: bot: I didn't get that, what can I help you with?

arsena commented 3 years ago

One problem with that is I'm not sure if Lex postText allows a blank string. Post content possibly allows a silent audio file. Can you confirm the postText behavior for me?

On Tue, Jun 1, 2021, 9:33 AM David Bell @.***> wrote:

Is it possible to get ChatPickle to allow a silent response from the user? Suggestion below passing to response with silence to bot.

bot: what can I help you with? user: bot: I didn't get that, what can I help you with?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/libertymutual/chatpickle/issues/42, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAGDQQUN6MFT4EDYT7Y3BNTTQTORVANCNFSM454V3VHA .

dtbell99 commented 3 years ago

It appears inputText must have a length of 1. The only thing I see that may be researched could be x-amz-lex:start-silence-threshold-ms and set it to zero and possibly x-amz-lex:end-silence-threshold-ms to also be zero. Then post to it with those and see if it responds with the expected result. Both of these are session attributes which can be passed in the postText json body.

arsena commented 3 years ago

This also might be possible if we implement voice... see this discussion starting here. https://github.com/libertymutual/chatpickle/issues/44

dtbell99 commented 3 years ago

I was not aware of the postContent functionality. May be an opportunity to tie into Polly to generate the audio and then post it and perhaps even save the utterances in a file so if you already have a "Yes" or a "Pay my Bill" it would map that utterance to a audio file "audiofile1.mp3" in a local json config so it only goes to polly each time you have a new utterance. This would not be efficient in a pipeline unless you had control of the storage but a really nifty innovation to this library.