[abcxjeb284]

Bit of a miss on why “serverless” (though author & I have different definitions), possibly because of language / toolset.

Lambda / Step Functions make for an easy combo for easily understandable units that you can use without a bunch of servers. Combine it with AWS CDK and you end up with a low boilerplate, auto-completeable environment (with types!) that’s actually easy to unit test locally!

I agree that yaml config and figuring out how to wire things up sucks. I’m open to Smalltalk being better.

But we’ve been using Lambda / Step for a few years now and it’s been highly robust, easy to understand, monitoring / logging essentially free and - best of all - we haven’t had to manage any server upgrades / patches / etc the whole time. (And super easy to spin up stack per user)

Have a queue of items to work on? Combine lambda + sqs (just a few lines with cdk) and now you have code that auto scales linearly with how many entries are in your queue.

We did write our own tool to handle boilerplate of linking python functions to their lambda handlers (but again this is the beauty of actually being able to use programming language to define your stack)