the worker works with vercel ai sdk and openai api. the issue here is I am using the vercel ai sdk a

the worker works with vercel ai sdk and openai api. the issue here is I am using the vercel ai sdk and after a few message with model, i get the error from worker tell that "cpu time exceeding". I am in the free plan. so there is 10ms limit. the issue here is before return the response to client, the vercel ai sdk doing some transform for that response, these transformation need cpu time to transform the response message from openai.

Tthanhle the worker works with vercel ai sdk and openai api. the issue here is I am using...

Isaac McFadyen•2/21/25, 6:03 PM

You could always use the OpenAI API directly, it's a standard HTTP API and they have reference documentation published that you can use

fetch

fetch

with: https://platform.openai.com/docs/api-reference/

NicoleLily•2/21/25, 6:30 PM

@Isaac McFadyen

NNicoleLily @Isaac McFadyen

Hard@Work•2/21/25, 6:31 PM

?pings

SuperHelpflare•2/21/25, 6:31 PM

Please do not ping community members for non-moderation reasons. Doing so will not solve your issue faster and will make people less likely to want to help you.

NicoleLily•2/21/25, 6:31 PM

@SuperHelpflare

Silvan•2/21/25, 6:42 PM

Hello everyone. i just saw the new vite-plugin-cloudflarevite-plugin-cloudflare

i current use sveltekit with worker assets

now i wonder. does this plugin do anything for me? do i understand correctly that my local dev will just get closer to actual cf infra?

custom•2/22/25, 2:15 AM

can i use disposable api in workers?

await using foo = new SomeAsyncDisposable()

await using foo = new SomeAsyncDisposable()

Ccustom can i use disposable api in workers? ```typescript await using foo = new SomeAsy...

Isaac McFadyen•2/22/25, 2:16 AM

Yes but it will be transformed at build time, and you must use a specific version of Wrangler: https://developers.cloudflare.com/workers/runtime-apis/rpc/lifecycle/#how-to-use-the-using-declaration-in-your-worker

Isaac McFadyen•2/22/25, 2:17 AM

This version of Wrangler will transpile using into direct calls to Symbol.dispose(), before running your code or deploying it to Cloudflare.

Walshy•2/22/25, 1:10 PM

Improvements on the way but yeah that is very heavy sampling

Walshy•2/22/25, 1:10 PM

cc: @Thomas Ankcorn

Benos•2/22/25, 4:29 PM

Question around Rate Limit bindings https://developers.cloudflare.com/workers/runtime-apis/bindings/rate-limit/ is there a way to add more than 1 to the rate limiter?
My use case is as follow:

Worker is called and forward to a backend service via HTTP (hosted outside of cloudflare) based on the "ClientID" (in the headers)
The backend calls another API and may get blocked
I know that the send API has a rate limit of around 150/10sec
I want to return 429 to as many requests as possible to not overwhelm the backend

Simple implementation is just to put the same rate limit
More advanced one is to increase the key more than once if I get a 429 from my backend.
I don see a way to do something like:
const { success } = await env.MY_RATE_LIMITER.limit({ key: pathname }, increment_num=20)
Should I just loop 20 times over env.MY_RATE_LIMITER.limit({ key: pathname }) without an await and return the 429?

Cloudflare Docs

Rate Limiting · Cloudflare Workers docs

Define rate limits and interact with them directly from your Cloudflare Worker

danboyle8637•2/22/25, 6:17 PM

In my app I want to send a request to another Worker using my service binding but I need to construct the request as a new request.

However when I build my new request -- new Request()new Request() I get a type error when passing the request to the env.BINDING.fetch(req)env.BINDING.fetch(req) because my request is a native fetch request and not a Cloudflare request.

How can I type my new request so I avoid the type error when passing it to the binding?

matt•2/23/25, 9:16 AM

hi everybody

i built a prototype of how you might get the new bun frontend build tools to work with cloudflare pages:
https://github.com/mattfysh/bunfun

matt•2/23/25, 9:21 AM

more info on bun's new release here: https://bun.sh/blog/bun-v1.2.3

Silvan•2/24/25, 2:20 AM

can someone recommend a validation library that works on cf? just had to learn the hard way that typebox uses eval

James•2/24/25, 3:22 AM

I use Zod in lots of workers and it’s been great

pandalambda2•2/24/25, 3:26 AM

what fine grain permission do you need to setup a workers route?

JJames I use Zod in lots of workers and it’s been great

Isaac McFadyen•2/24/25, 4:06 AM

^ I didn't know if Zod was Workers-compatible or not but it sounds like it is, and yeah, Zod is IMO the best

Niveth•2/24/25, 9:25 AM

Cloudflare smart routing placements seams incorrect?

I have database hosted on ap-south-1 (BOM) and I have my API deployed on cloudflare workers with smart placement enabled. When I send a request from Delhi, India, I expect the request to be routed to either DEL datacenter or BOM (Preferred) datacenter but for some reason it is getting routed to FRA (Germany) and sometimes it is getting routed to SIN (Singapore).

Because of this my I am having huge latency between my database and api which makes our whole app a lot slower.

Any tips on how to fix this ?

IIsaac McFadyen ^ I didn't know if Zod was Workers-compatible or not but it sounds like it is, a...

Silvan•2/24/25, 11:52 AM

i really dislike zod so yea. either i could use something like this https://www.npmjs.com/package/@cfworker/json-schema

or i put it in a buildstep i guess

npm

@cfworker/json-schema

A JSON schema validator that will run on Cloudflare workers. Supports drafts 4, 7, 2019-09, and 2020-12.. Latest version: 4.1.1, last published: 24 days ago. Start using @cfworker/json-schema in your project by running npm i @cfworker/json-schemanpm i @cfworker/json-schema. There are 31 other projects in the npm registry using @cfworker/json-schema.

NNiveth Cloudflare smart routing placements seams incorrect? I have database hosted on...

Niveth•2/24/25, 2:59 PM

Finally solved it, you will have to get cloudflare pro account for a custom domain and set up same custom domain for the worker.

Basically without pro account cloudflare workers are not much useful in India

NNiveth Finally solved it, you will have to get cloudflare pro account for a custom doma...

sathoro•2/24/25, 3:37 PM

"useless"

Victor•2/24/25, 4:38 PM

Looking at gradual deployments, is there a way to specify a minimum version by date?

Seems crazy that there's no such header for this very common use case: deploy workers + app the same time, gradual rollout for older versions, newer versions must use workers past their build date

mihaaai•2/24/25, 4:45 PM

How to expose the Cloudflare environment form wrangler to the front-end (client) side of a full stack app with Vite?
I need to export the environment name previewpreview or productionproduction (the only environments accepted by wrangler for a pages project), but I don't know how to intercept this at build-time with the Github Integration of a SolidStart app, build with Vite.
Ideally I only need to export a URI for each environment this being the main purpose, the previewpreview app needs to make some API request from the client to an endpoint different form the productionproduction, and since Environment Bindings are only exposed from the context object of the workers I can't reference the correct URI for different remote environments
previewpreview is gonna be my staging while productionproduction is of course the prod.

Ruxorly•2/24/25, 5:23 PM

when will outbound TCP to cloudflare ip ranges be re-enabled?

Ssathoro "useless"

Niveth•2/24/25, 6:11 PM

I think my reaction was bit strong. I wasted one whole day going through community forums to figure out the issue. I really wish cloudflare addressed in their documentation somewhere so we can just buy pro account in advance.

Other than this one particular issue, I had really good experience with all other cloudflare offerings like queues and durable objects.

Brett Willis•2/24/25, 8:01 PM

So stoked about the new Images bindings API! Got some questions though:
In this example, doesn't the env.IMAGES.info(stream)env.IMAGES.info(stream) call consume the ReadableStreamReadableStream and then env.IMAGES.input(stream)env.IMAGES.input(stream) try to reuse the stream that is already consumed? Is there some magic going on there?
Can I call env.IMAGES.info(stream)env.IMAGES.info(stream) on a stream just to verify info and then pipe the same already-consumed (?) stream to R2 without any transformations? Or do I need to tee the stream?
Also how are these transformations priced? Is .info(stream).info(stream) free? Do I assume that each call to <some transformations>.output()<some transformations>.output() triggers one "unique transformation" quota usage? E.g. 5k free

.output()

.output()

calls per month then $0.5 per 1k thereafter?

Keebs•2/24/25, 8:26 PM

is there a way to deploy pages & workers at the same time? like if I have pages using 5 workers... can I just have it all in one package?

Keebs•2/24/25, 8:27 PM

oh nevermind

coliquet•2/24/25, 8:35 PM

I know this is a really stupid question, but I'm legit so lost. So many forums and documentations mention or are telling me to go to workers and pages but I can't find it for the life of me. I found a guide that says you can find it at your account home but the only thing I see that says workers or pages is "Workers for Platforms"

Ccoliquet I know this is a really stupid question, but I'm legit so lost. So many forums a...

Hard@Work•2/24/25, 8:37 PM

This link should take you there, after selecting your account(if you have multiple): https://dash.cloudflare.com/?to=/:account/workers-and-pages

coliquet•2/24/25, 8:38 PM

Thank you! I realize I should've posted in one of the help channels

dave•2/24/25, 9:16 PM

Does the cache not work in Cron handlers?

Ddave Does the cache not work in Cron handlers?

Hard@Work•2/24/25, 9:34 PM

No, since they run under your workers.devworkers.dev, which has Cache disabled

HHard@Work No, since they run under your `workers.dev`, which has Cache disabled

dave•2/24/25, 10:01 PM

ah.... No way around that I guess?

Ddave ah.... No way around that I guess?

Hard@Work•2/24/25, 10:25 PM

What are you trying to do?

HHard@Work What are you trying to do?

dave•2/24/25, 10:26 PM

Have caching on my

fetch

fetch

's.

Ddave Have caching on my `fetch`'s.

Hard@Work•2/24/25, 10:26 PM

Between multiple runs?

HHard@Work Between multiple runs?

dave•2/24/25, 10:27 PM

Yeah, would be nice.

Ddave Yeah, would be nice.

Hard@Work•2/24/25, 10:28 PM

Yeah. I would note there that you aren't guaranteed to be running in the same colo twice in a row, so it may not make a huge diff. That's said, you can't do this at the moment

HHard@Work Yeah. I would note there that you aren't guaranteed to be running in the same co...

dave•2/24/25, 10:28 PM

True, but I've noticed they tend to stick in a colo for awhile.

1984 Ford Laser•2/24/25, 11:12 PM

This new Images binding looks great! Definitely will simplify multiple parts of my graphics workflow

One question, is there a limit to the number of

.draw()

.draw()

commands that can be done in a single request? Through the existing

draw: []

draw: []

method, I have been limited to ~70 layers due to 16KB header limit, hoping this will be able to get around that but not sure if the binding is just doing a

fetch()

fetch()

underneath

11984 Ford Laser This new Images binding looks great! Definitely will simplify multiple parts of ...

Isaac McFadyen•2/24/25, 11:14 PM

Most bindings are doing fetches underneath, but whether it uses the same format (passed in the header) will depend on the implementation.

IIsaac McFadyen Most bindings are doing fetches underneath, but whether it uses the same format ...

1984 Ford Laser•2/24/25, 11:15 PM

All good, I have a batching method that works fine (draw 25 layers, save result, draw next 25 layers on saved result), but this should still simplify that setup a lot

IIsaac McFadyen Most bindings are doing fetches underneath, but whether it uses the same format ...

Hard@Work•2/24/25, 11:29 PM

It appends it as JSON in a FormData payload

HHard@Work It appends it as JSON in a FormData payload

Isaac McFadyen•2/24/25, 11:30 PM

Good to know, so the 16KB limit wouldn't apply to that then @1984 Ford Laser ^

the worker works with vercel ai sdk and openai api. the issue here is I am using the vercel ai sdk a

Similar Threads

Similar Threads

Similar Threads