- Newest
- Most votes
- Most comments
The Start Streaming block will start sending the audio to Kinesis Video Stream (does not send it to S3 or anywhere else), different from Kinesis Data Stream. What you will need to build is a consumer of these RAW live audio streams and construct the wav file yourself, here is an example:
https://github.com/amazon-connect/amazon-connect-realtime-transcription
The example contact flow shown disconnect the calls and does not send to a queue, so what you can do is modify the contact flow to send the call to a queue and eventually to an agent, the entire call will be streamed and you will be able to capture it via KVS. Regular Connect call recording will continue to work (only between caller and agent portion)
I've never seen a Kinesis actually be used to record Lex interactions, I suspect it's not possible. Kinesis is mainly used to build a "voicemail" functionality or to allow the caller to record feedback like for a survey. Check this link out https://s3.amazonaws.com/solutions-reference/voicemail-for-amazon-connect/latest/voicemail-for-amazon-connect.pdf and you'll get a walkthrough on how it can be used to capture caller audio.
david
That's not true. Here https://docs.aws.amazon.com/connect/latest/adminguide/set-up-recordings.html it says: Tip To record the customer when they are not connected to an agent, for example, if all of their interaction is with a Lex bot, use media streaming. And here: https://docs.aws.amazon.com/connect/latest/adminguide/customer-voice-streams.html it says: The customer audio streams also include interactions with an Amazon Lex bot, if you're using one in your flow.
Relevant content
- asked 6 months ago
- asked 2 months ago
- AWS OFFICIALUpdated 2 years ago
- AWS OFFICIALUpdated a year ago
- AWS OFFICIALUpdated 3 months ago
- AWS OFFICIALUpdated 2 years ago
Thanks for helping. But what I read is that the data was going to be retained by 5 minutes, I was expecting to process it once Stop media streaming block was executed using a lambda function as trigger. What you are suggesting is way too complicated for me, I will just remove this feature from my app.