After the user invokes an app on Alexa , Is there a way to get the query as a voice stream/audio file of a user? Through alexa I want to send the stream to a webservice/lambda that the invoked app will call and analyze the intent there.
We have some proprietary code that we want to use for analyzing intent hence we cant do it on the alexa side Since I am sending the query after the user has invoked the app and through the app there are no privacy concerns(hopefully)
Thanks
No, that is not possible, and I don't think it will be.
Echo devices connect to Amazon only, and Amazon uses Lex (which is also available via AWS) to parse speech files. As a skill developer, you will only receive the parsed results: intent, slots - and maybe, when Amazon implements user differentiation, an anonymous ID for the speaker.
There is no way to access the original speech audio in your skill. As every file is also used by Amazon to train their speech recognition, I doubt they will open their ecosystem accordingly.
Only option I see currently: build your own Echo with e.g. a Raspberry Pi, then you have full control. But you can't leverage the install base of Echo.
Same applies to Google Home and Microsoft Cortana, so it's not just Amazon.
