Design considerations - AI Powered Speech Analytics for Amazon Connect

Design considerations

Deploy an alternative CloudFormation stack

By default, this solution deploys with AWS Lambda to process the audio from customer calls. However, AWS Lambda has a limitation where audio processing stops if a call lasts longer than 15 minutes. To avoid this limitation, you can deploy this solution using Amazon SQS and Amazon ECS that process the audio for up to one hour. To launch this alternative setup, you must change the ComputeType parameter from the default Lambda setting to Fargate. For more information about this parameter, refer to Step 1. Launch the stack in the Automated deployment section.


        AI Powered Speech Analytics for Amazon Connect architecture on AWS using AWS Fargate

Figure 3: AI Powered Speech Analytics for Amazon Connect architecture on AWS using AWS Fargate

This alternative deployment deploys AWS CodeBuild to create a Docker image for the application and uploads it to an Amazon ECR repository to allow an AWS Fargate task to locate it. An Amazon ECS cluster is created along with an associated Fargate task definition to run the container’s logic.

When a customer calls into your Amazon Connect call center, their call progresses through the contact flow which invokes a Lambda function to emit a message to the SQS queue. Customer audio is streamed in real time through Amazon Kinesis Video Streams.

The Fargate task polls for SQS messages, then consumes the audio stream and uses Amazon Transcribe to convert the audio into text. The Fargate task then stores the transcript segments and contact ID in an Amazon DynamoDB table.

This alternative deployment uses two CloudWatch alarms to check for the number of items in the queue, and to scale the number of ECS containers. One alarm scales up the number of ECS containers while the other alarm scales the number of containers down.

Regional deployment

AI Powered Speech Analytics for Amazon Connect uses Amazon Connect, Amazon Translate, Amazon Comprehend, and Amazon Kinesis Video Streams, which are currently available in specific AWS Regions only. Therefore, you must launch this solution in an AWS Region where these services are available. For the most current service availability by Region, refer to the AWS Regional Services List.

Note that Amazon Transcribe Streaming is not supported in every region where Amazon Transcribe is supported. Refer to Amazon Transcribe endpoints and quotas for a list of AWS Regions where Amazon Transcribe and Amazon Transcribe Streaming are available.

Updating the solution

If you have previously deployed the solution, you must update the solution's CloudFormation stack to get the latest version of the solution's framework. For details, refer to Update the stack.