J@anthony the limits and allowances are pretty generous on the paid plan though, especially if it's a cron trigger or queue consumer (15 minutes)
J@anthony are you bumping up against some limits?
PI naively thought I could share the logic from my nextjs app with my cloudflare workers but the environment variables is causing quite a bit of issues. Even when polyfilling with
import * as process from 'node:process';
, whenever I use process.env, the nextjs app doesn't understand "node:process". Is it common to simply duplicate the logic for the cloudflare worker? Even my database calls requires a separate env and it's a bit too integrated. Is it common for folks to duplicate their code logic for CF workers or are people commonly able to reuse logic from their app?PIs it common to declare something like globalThis.process = process (imported from node:process) when using cloudflare workers for nextjs and cloudflare worker compatability?
LHi everyone ! Is there a way to send a fetch request from the Cloudflare worker without waiting the response ? The API that I try to contact is just there to receive data and then will do a lot of operations that take some time and I want to close my connection without waiting for the reponse :)
Cheers !
Cheers !
Cctx.waitUntil is your friend, hand it a promise and it'll let the response return but not kill the execution
ctx.waitUntil(fetch(..))
ctx.waitUntil(fetch(..))
CCloudflare Docs
The Context API in Cloudflare Workers, including waitUntil and passThroughOnException.
LWe've been exploring waitUntil recently. I'll let you know if this is the magicial solution we needed @Chaika :)
Wumm what did i do wrong
APretty similar as to what i've read here. Ingesting PDF's via fetch in context of http request. Want to parse it for the text using unpdf, so i can use it in Workers AI/LLM. Works fine code-wise but explodes in prod because of timeout/memory issues.
Works for small pdf's but reality is i can have some larger ones as well which does not seem possible. Streaming is not an option afaik forthese operations...
Works for small pdf's but reality is i can have some larger ones as well which does not seem possible. Streaming is not an option afaik forthese operations...
JYeah, pdf is a pig of a format. My only recommendations are, if you're hitting timeouts, make the process asynchronous by storing the pdf in R2 and putting a message on a queue. Then you'll have 15 minutes to process it.
If you're hitting memory limits, one possibility could be too find a way to split the pdf into pages without loading the whole thing into memory (I don't know if this is actually possible given the format).. maybe using WASM
If you're hitting memory limits, one possibility could be too find a way to split the pdf into pages without loading the whole thing into memory (I don't know if this is actually possible given the format).. maybe using WASM
J@anthony this looks interesting.. https://dev.to/wcchoi/browser-side-pdf-processing-with-go-and-webassembly-13hn
HIf a call doesn’t immediately start? It doesn’t throw an error, it just pauses it until a slot frees up
J@anthony ooo.. looks like mupdf can handle streams.. https://github.com/ArtifexSoftware/mupdf.js/blob/master/examples/tools/split-pages.js
DWhat's the fastest/highest perf way to validate an incoming request for valid JSON?
DHmm, that’s gonna require buffering the whole thing in memory first though. 
DI don’t think CF would even notice at that tiny volume.
DYou can do a fetch with zone caching enabled too iirc.
AInteresting, going to look into this!
JAWhat range of pdf sizes were you having trouble with btw?
JOnly 2MB, so didn't have memory issues, but nearly 200 pages, so took 5 seconds or so.. but fast enough that I could just process it in response to a request
AHow to download and display images from Workers R2 binding on a page using Javascript?
Shey @Unsmart | Tech debt @MetalGear is this still work in progress?
HHello! I was looking at this example https://github.com/cloudflare/js-rpc-and-entrypoints-demo/tree/main/public-admin-api-interfaces about using a dedicated worker for DB interactions using RPC for inter-workers calls.
From my understanding of the docs, RPC calls are usually executed in the same machine.
However, workers with D1 attached are usually executed closer to D1.
In this case what would happen? Will this RPC call happen in the same machine with the trip to D1 location or the RPC call will happen via network?
Example:
worker-db with a D1 attached in EU
worker-front with service bind to worker-db
Then worker-front receives a request from US, which usually happens closer to the user.
What will happen then?
1 - worker-front and worker-db executed closer to user (US) and network trip to D1 db?
2 - worker-front execution in US and RPC call via network with worker-db running closer to DB (EU)?
3 - Both workers executed closer to DB (EU)?
4 - Something else?
From my understanding of the docs, RPC calls are usually executed in the same machine.
However, workers with D1 attached are usually executed closer to D1.
In this case what would happen? Will this RPC call happen in the same machine with the trip to D1 location or the RPC call will happen via network?
Example:
worker-db with a D1 attached in EU
worker-front with service bind to worker-db
Then worker-front receives a request from US, which usually happens closer to the user.
What will happen then?
1 - worker-front and worker-db executed closer to user (US) and network trip to D1 db?
2 - worker-front execution in US and RPC call via network with worker-db running closer to DB (EU)?
3 - Both workers executed closer to DB (EU)?
4 - Something else?
GitHub
Contribute to cloudflare/js-rpc-and-entrypoints-demo development by creating an account on GitHub.
NHey guys!
Might be a dumb question but, is the Cloudflare workers cache API suitable for credential and authorization policy caching?
Would it be secure to store this kinds of sensitive information in a custom cache instance (cache.open)?
Are cache instances guaranteed to be private inside a CF account?
Might be a dumb question but, is the Cloudflare workers cache API suitable for credential and authorization policy caching?
Would it be secure to store this kinds of sensitive information in a custom cache instance (cache.open)?
Are cache instances guaranteed to be private inside a CF account?
HThey are guaranteed to only be accessible within your account. You can even use a fake URL like
https://mystore/somepersonNthat was exactly my main worry, i, currently using something to the lines of https://<tenantid>.policyengine/<role>
i was worried if collisions with other accounts were a risk
i was worried if collisions with other accounts were a risk
Nthank you for the clarification
LCan I call multiple user workers from a dynamic dispatch worker in a single request?
NAbout the cache purge situation, is there really no way to purge a single URL outside the enterprise plan?,,
Thats inviabilizes the idea.. if so i might need to go to KV or Redis instead
I was thinking of using it as a Cedar policy cache, i managed to run the recently open-sourced cedar policy engine (the AWS permissions engine) inside a rust worker, and im trying to build a low cost api gateway alternative with CEDAR policy resource management suport. adding the requirement of one or more kv requests per api call seems a shame while having an amazing cache api, but an authorization service without near instant policies updating seems.. wrong
Thats inviabilizes the idea.. if so i might need to go to KV or Redis instead
I was thinking of using it as a Cedar policy cache, i managed to run the recently open-sourced cedar policy engine (the AWS permissions engine) inside a rust worker, and im trying to build a low cost api gateway alternative with CEDAR policy resource management suport. adding the requirement of one or more kv requests per api call seems a shame while having an amazing cache api, but an authorization service without near instant policies updating seems.. wrong
HFor Zones on Free/Pro/Business plan, you may purge up to 30 URLs in one API call
HThough this only applies to
caches.default iircNOh.. thats.. ok i guess?
Would it be ok to save the cache in the default cache while using a fictional url like we talked about above?
Would it be ok to save the cache in the default cache while using a fictional url like we talked about above?
LAnd if so, how would that be billed?
HIt should be fine. You could even use a real URL that isn't on your account, like
google.comNIf that URL never comes to actually exist, and be publically hosted in my CF account, theres no risk of the contents being exposed, is that correct?
NIf thats not too much to ask, would you be able to provide me with a link to the docs for that API call or SDK? On the cache documentation i only found the prefix based cache purge (marked as Enterprise)
HInteract with Cloudflare's products and services via the Cloudflare API
NThanks!
Ill try it out, if i have good results (and performance) ill probably open source my cedar-policy-on-cf repo
Ill try it out, if i have good results (and performance) ill probably open source my cedar-policy-on-cf repo
NRight now im getting 3ms for policy lookup, parsing and allowed/denied outcome on a 20 entities, 10 principals setup, seems amazingly quick! With added JWT parsing (and maybe caching with TTL equal to token exp) im thinking ican keep it under 25ms total, which blows AWS API Gateway out of the water!
HI’d say no. Some places even just use 6 digits. Though, you could probably go much longer if you made it a link
Sis there a way to decrease number of requests for workers , as each visit to my website is hitting the worker hundreds of time , for each path , and for each request it count 1 request
?
?
Seach visit cost me more than 1000 request worker , I just need the worker to work 1 time for each ip address