Did you enable the `nodejs_compat` flag?
Did you enable the
nodejs_compat flag?
HWhat you are showing is
--node-compat, which is the old polyfill methodHIt has support for less stuff
H
DThat fixes my example code but I'm working with a third party library which no longer works when using
nodejs_compatcrypto instead of node:crypto. Is there some way I can make this third party library use the polyfill modules?
HHRename
crypto to node:crypto, then mark it as externalDThat might work, but it seems a little weird to me that this is the best way to do it when
--node-compat worked out of the box. Is this really the way to go when using any third party library that uses node modules?HMaybe? I generally just avoid deps that require Node, but I know that isn't feasible for everyone
HDon't remember the exact reasoning for it, but iirc it was something about Node's APIs not being JS-standard, so they wanted to scope them
HJust like how non-standard Cloudflare APIs that require an import are scoped
cloudflare:
KGiven there’s 4,521 dependents on the “crypto” lib on npm (which is built into Node now, yes) still 
KTIL they literally nuked it other than a deprecation README
KAnyways, there was work in Wrangler to make it automatically turn crypto -> node:crypto (and others) for you but that isn't done yet
DHow do I supply arguments like this to esbuild through wrangler?
DI'm not very familiar with TS build tools at all 
HYou have to install
esbuild, then run the build yourselfHThere are like 3 things you can configure on esbuild when using the
wranglerThat sounds like it would need a lot of config to get it to work with cloudflare
DIt seems the library I'm using does have a browser version you can build yourself. I'll look into how to integrate that
Aanyone else not able to deploy? sees like wangler login isn't working. Gate way error 504
DOkay, this doesn't quite seem to work either. Is it possible it's trying to access browser APIs that CF Workers don't have access to?
DI'm not getting any build errors this time, it just doesn't work and throws warnings and tries to reconnect
KWhat package are you trying to use?
DI'm using gramjs, a Telegram client. It has both a node and browser version
DCF Workers might not be the best way to do this but I need something to automatically send a couple of telegram messages once per day
Aworking again
DDoes wasm binaries bundled using
Because my worker.js size is 940KiB and overall size including wasm is 2.3MiB after bundling but it throws the error
wrangler deployBecause my worker.js size is 940KiB and overall size including wasm is 2.3MiB after bundling but it throws the error
workers.api.error.script_too_largeKYes
Amhhh... if I am doing wrangler dev (and I dont do await on a htmlrewrite transform), it doesnt work. But it works deployed when doing return rewriter.transform(response);
AI should just do await then right?
AIn the docs there is no await so thats why I was confused!
KYou can return a Promise from an async function (i.e the
fetchyeah, I do that... but seems like when using locally (so without --remote) the htmlrewriter isn't running. Le tme create a sample project to show it
Aneed to find a nice way to create apis that can be consumed by workers. openapi seems so heavy these days.. tho maybe hono makes it nice
Ausing stuff like zodios/trpc and messing around with mono repos also an option i guess
BI'm having a hard time grasping the worker concept of environment variables. I'm using hono and I want to create a variable outside hono 'c.env' context. I don't understand Where this 'env' comes from.
Bfor example:
app.use('/input/*', bearerAuth({ TOKEN })); I would normally use 'process.env.TOKEN'
HYou would do something like
BMy hero! What it I just want to access them outside of the "app." context?
HLike, in the global scope? Is this for initializing a client or something?
Byes
HThen you would initialize it within the
app scope, then store it in a globalHYou could also use an AsyncLocalStorage to have it accessible globally(but that requires a compat flag)
HBut note, env vars are only ever available in the context of handling an incoming event
HFor example, you can't just pull a secret on startup
Bok, I guess I just have to repeat initialization code for each endpoint. I'm using Langhian and have to initialize the embeddings and vector store for every GET and POST. Is there a best practice?
HYou can have a single global variable to store them in, so they only get initialized once
HThough note too that isolates aren’t reused very often, so it might be over optimization
HGenerally, they are only reused by the same client, in a limited period of time afterward
