Hello everyone! I am so incredibly excited to share that we are building a real-time sentiment analysis bridge using AWS Comprehend for our Genesys Cloud Web Messaging sessions! The goal is to provide our supervisors with a live ‘Sentiment Score’ dashboard! We are using an AWS Lambda function triggered by the Notification API (v2.conversations.messages.{id}). However, I am noticing a strange issue where the message events are sometimes arriving out of order at our Lambda! This makes the sentiment calculation look very jumpy. Has anyone else found a clever way to re-order these events before passing them to Comprehend? I am so eager to make this work perfectly!
Hey! I am doing something similar but for an ETL historical data migration. The ordering problem is a total nightmare because the Notification API does not guarantee delivery sequence. To fix this, you should check the eventTime or metadata.timestamp inside each message segment. Instead of processing the messages as they arrive, we dump them into an Amazon Kinesis stream first. Kinesis allows you to buffer the messages for a few seconds and then use a consumer to sort them by timestamp before you hit your sentiment engine. It is a bit of a workaround, but it is the only way to keep the context straight!
Good afternoon. I manage a large fleet of on-prem Edge servers and I have observed this exact same behavior with high-concurrency messaging sessions. The out-of-order delivery is a side effect of the Genesys Cloud distributed event architecture. The platform prioritizes throughput over strict sequencing for performance reasons. If you are building an enterprise-grade sentiment dashboard, you must implement a state-aware buffer. You cannot rely on the arrival order. Utilizing a sequence number from the messageMetadata is the standard protocol for reassembling these interactions correctly.