Zander here, part of the team working on Bytewax. Much of the magic of Timely is passed through to the Python library, but there are a few features that we thought would make it easier to build with and run in production.
- We deepened the integration between inputs and output sources starting with Kafka/Redpanda.
- We have added state recovery backed by either SQLite or by a Kafka API compatible platform like Redpanda or Kafka.
- Built a deployment tool called Waxctl that makes it easier to run dataflows across pods on a kuberentes cluster and deploy to remote compute cloud VMs like EC2.
- Provide some helpers for things like windowing and ordering inputs. These are still evolving
We are always looking for feedback and for which features to focus on next!
