From the cloud to the factory – Temporal used in manufacturing applications

Bright Machines is a startup working on software-defined manufacturing and we have developed our own orchestration systems for industrial automation. However, we see that switching most of our non-realtime process controls to Temporal will significantly reduce our non-differentiating engineering effort on orchestration, including support and training. This talk presents the journey to running a manufacturing line on Temporal.

Using Temporal enables us to get most of the features we need without implementing them, and without degrading performance. We will walk through these in the talk:

• Systematic metrics, with percentiles, for every step of the manufacturing processes, out of the box.
• Component traceability is implemented just by returning the assemblies' bill of material with serial numbers and measured characteristics at every step of the manufacturing. As such, these characteristics are logged in the workflow history.
• Data retention to support future recalls or other quality investigations: the workflow archival feature provides a base coverage for what we need, out of the box.
• Unify on one “lingua franca” language for us, Python, without having to get rid of other components in other languages, and without precluding developing new components in other languages.
• Logical bugs in the process design can be systematically inspected and troubleshooted at the mothership, away from the production lines, by replaying the workflow history in a development environment.