Answering this post by showing an AWS Lambda example.
Clone repo and run
sh build.sh
This creates a dist/lambda.zip.
Deploy dist/lambda.zip to AWS. I'm sure you can do this via the command line somehow. I just created the AWS Lambda in their web interface and uploaded it manually for testing. I know next to nothing about AWS, just want to demonstrate the CLJS bits.
For this example I had to bump the default runtime timeout. 3sec didn't seem enough. The @mozilla/readability part seems to take about 5sec to run. 10sec timeout should be plenty. Don't know why it is so slow, it isn't in the CLJS bits.
Let's go over what build.sh does. You can use anything to perform these steps, but this seemed easiest to explain.
rm -rf dist
mkdir dist
Create an empty dist directory just so we don't copy files we don't need.
npx shadow-cljs release lambda --config-merge '{:output-to "dist/index.js"}'
Tell shadow-cljs to make a release build for the :lambda build. Using config-merge here, so it outputs the file directly to the dist folder. Could do this via the build config directly too.
cp package.json package-lock.json dist
cd dist
npm install --omit=dev
Ensure that all dependencies we may need are installed in the dist directory itself. the --omit=dev will not install devDependencies. Meaning that shadow-cljs itself won't be in the dist dir since it is not required. Makes the zip significantly smaller.
rm package-lock.json
zip -r lambda.zip .
Finally, create the zip file by zipping up the entire dist folder. You may also delete any additional files you do not need. So, just deleting the lock file as an example. It won't hurt anything if you keep it.
The build result here is a 2.5mb zip file, which is mostly due to node_modules files. The CLJS output itself is 22kb zipped, 98kb unzipped.
Deployment via s3 buckets is also an option, but I'll leave you to figure out AWS specifics on your own.
For local development you can run npx shadow-cljs watch lamba or npx shadow-cljs server and using the http://localhost:9630/builds/lambda UI to start/stop the watch.
Executing node test.js runs the local output. Of course that is going to get limited quick. I don't know what options AWS provides for actually running lambda locally. You'll probably want to adjust the event parameter in test.js at some point. You won't get a null in a real lambda after all.
If the server is running the build.sh will also be significantly faster, since it doesn't need to start a new CLJS instance.
If you want a REPL use npx shadow-cljs node-repl, as the :lambda build has it disabled. Didn't want to have the node process linger after calling the handler via test.js.