As a follow up to this thread: https://discord.com/channels/905144809105260605/1113936455954346005/1141738248805298218 Would Medplum consider using a logging library again? Or exposing an option to provide a custom logger? Or some 3rd thing I'm not thinking of (other than a fork)? We're trying to do distributed tracing through our API to Medplum (with Datadog's dd-trace). The stack traces work perfectly, it's super cool. But one feature we are missing is connecting logs to traces. We need to be able to emit a trace id in log messages in each service. dd-trace automatically patches winston, pino, et al - but console.log requires manually writing the trace id. So far we've come up with this truly awful Dockerfile option where we patch console.log:

FROM medplum/medplum-server:latest as medplum
RUN npm install --save dd-trace
RUN echo "const tracer=require('dd-trace'),formats=require('dd-trace/ext/formats'),originalConsoleLog=console.log;console.log=(...args)=>{const span=tracer.scope().active(),record={};if(span){tracer.inject(span.context(),formats.LOG,record);}if(args.length===1){try{let parsedArg=JSON.parse(args[0]);if(typeof parsedArg==='object'&&parsedArg!==null){args[0]=JSON.stringify({...parsedArg,...record});}}catch(e){}}originalConsoleLog(...args);};tracer.init();" > tracer.js
ENTRYPOINT [ "node", "--require", "./tracer.js", "packages/server/dist/index.js", "env:" ]

@joshua_kelly this is something that's top of mind to us. We've written up a github discussion with our current thoughts here https://github.com/medplum/medplum/discussions/2759

I'm pretty sure yes, I'm almost certain that we can control the trace id header that we send in our request