Cloudflare Developers•11mo ago

https://developers.cloudflare.com/workers-ai/models/

Original message was deleted

Isaac McFadyen•2/21/25, 1:03 PM

?crossposting

SuperHelpflare•2/21/25, 1:03 PM

Please do not post your question in multiple channels/post it multiple times per the rules at #

welcome-and-rules. It creates confusion for people trying to help you and doesn't get your issue or question solved any faster.

11984 Ford Laser https://developers.cloudflare.com/workers-ai/models/

thanhle•2/21/25, 1:07 PM

i mean chat api from open ai. i understand we have worker AI. But it only for open source model and they are not good when compared to open ai or anthropic.

strainix•2/21/25, 1:12 PM

Can someone confirm that regional services are only available for enteprise customers?

strainix•2/21/25, 1:17 PM

Sorry, I found it. Ent only.

11984 Ford Laser You need to be moderating this content where possible. Many jurisdictions (Austr...

sodepr•2/21/25, 1:43 PM

It's hard; I cannot manage all the content that is uploaded by users.

11984 Ford Laser You need to be moderating this content where possible. Many jurisdictions (Austr...

sodepr•2/21/25, 1:45 PM

Moreover, some content is legal in this country but illegal in others.

Ssodepr Moreover, some content is legal in this country but illegal in others.

strainix•2/21/25, 2:21 PM

Well countries usually don't care if it's legal in other countries, you're providing content there which you are responsible for.

Tthanhle i mean chat api from open ai. i understand we have worker AI. But it only for op...

blindChicken•2/21/25, 2:40 PM

You don't need workerAI for openAI you just send requests to the openAI api. As to whether or not open source models are as good or not is open for debate. The best model will depend mostly on your use case.

Tthanhle i mean chat api from open ai. i understand we have worker AI. But it only for op...

dan•2/21/25, 4:25 PM

he might be meaning https://developers.cloudflare.com/ai-gateway/integrations/vercel-ai-sdk/

dan•2/21/25, 4:26 PM

you can't use workers AI with openai (doesn't make sense), you can use the ai gateway in front of it tho

thanhle•2/21/25, 5:17 PM

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-cloudflare

vite-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-schema

npm 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

preview

preview

production

production

(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

preview

preview

app needs to make some API request from the client to an endpoint different form the

production

production

, 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

preview

preview

is gonna be my staging while

production

production

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

ReadableStream

ReadableStream

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"