block-stream VS serverless-image-handler

What's probably got your shorts in a knot is the contents of the node_modules directory. That directory is a cache for mode.js modules so the tutorial can function standalone without having to install anything else. Treat it as you would object code. I pulled a half-dozen of those modules at random and went to their repositories. Every one of them had unit tests: block-stream, graceful-fs, lowercase-keys, minimist, pify and safe-buffer. Nobody sane ships unit tests with their executables.

AWS Suggests this if you want to do regular lambda + CF cache and they say it is production quality: https://github.com/aws-solutions/serverless-image-handler

I use imgix to resize images on the fly, which I recommend, but given the large amount of images you have it may not be cost effective. You could look into running your own: https://github.com/aws-solutions/serverless-image-handler

It looks like there’s a ‘standard service’ for video media: https://aws.amazon.com/elastictranscoder/ but actually for jpg you might need to do something a bit more bespoke. You could deploy a ‘serverless’ lambda function to do this conversion and write the S3 urls to a database after: https://github.com/awslabs/serverless-image-handler. Looks like it uses SharpJS: https://sharp.pixelplumbing.com/. In the past I’ve used ImageMagick on compute instances: https://imagemagick.org/script/index.php