You must configure the number of shards for each Kinesis Stream to support your anticipated call volume. The "correct" number of shards depends upon your call volume measured in calls per second and, to a lesser extent, on the length of your calls, since longer calls lead to larger JSON output documents. For typical Contact Center calls, a Kinesis stream with one shard is sufficient to handle approximately 25 calls per second.
VoiceBase does not recommend making any changes to the security settings of the API Gateway. The standard configuration should work fine.
VoiceBase has two output options. One output option is to place the JSON results on a new Kinesis stream. The second option is to place the JSON results in a Kinesis Firehose stream where the JSON results will be stored in an S3 bucket.
The first Lambda gets the notification about a call recording, VoiceBase configuration parameters and call metadata including the audio location. It creates a VB request by creating a pre-assigned URL to the recording and sends the trust to the VoiceBase API. Then VoiceBase uses this URL to pull the call recording.
No. The Lambdas will be created through the CloudFormation process.
The base integration installed via CloudFormation does not require any scripting. If you choose to create Custom Vocabulary groups, then additional work will be necessary. These will need to be created via the API. Full API documentation may be accessed here.
No additional credential configuration is required for VoiceBase to access the S3 bucket.
A sample payload may be viewed here.
Yes. The VoiceBase integration will retain the Amazon Connect metadata that is associated with the call.
No. The first Lambda only reads the CTR data from the Kinesis stream. It does not make any changes.
The Amazon Connect calls are automatically recorded in stereo. The high definition voice enables VoiceBase to provide a highly accurate speech-to-text transcription with speaker separation.