What is your opinion on modeling a request and marking set of rules on parameters so that it never hits your Lambda when the request is not valid to begin with? I see that your Lambdas will be called less, and that is good for pricing and throttling/concurrency. On the other hand, the Lambda pricing is very cheap, and if you dont have a heavy load it seems like you dont need it. I see the downside to this as that it will be more difficult to identify what part was invalid in the request. Because when it hits Lambda, you can log it to the CloudWatch. What do you think?

We've considered adding schema/models to the inbound API Requests to do pre-validation but never saw quite enough value in it to actually proceed and do it. I think if you had an app with lots of different complex payloads and a lot of traffic, it might make sense

Yeah, in our experience, if it’s just ur frontend that’s calling ur backend, you shouldn’t have much invalid request.

You tempted to move the SST-side @Ömer Toraman 😉