That's for Whisper Turbo, one of the Workers AI engineers confirmed a while back (they don't have no

That's for Whisper Turbo, one of the Workers AI engineers confirmed a while back (they don't have non-Turbo)

beelzabub•2/27/25, 11:52 PM

Did anyone figure out how to get workers AI to run with vitest locally? Getting the same issue as workers-ai

beelzabub•2/27/25, 11:52 PM

workerd/server/workerd-api.c++:759: error: wrapped binding module can't be resolved (internal modules only); moduleName = miniflare-internal:wrapped:__WRANGLER_EXTERNAL_AI_WORKER
workerd/jsg/util.c++:331: error: e = workerd/server/workerd-api.c++:789: failed: expected !value.IsEmpty(); global did not produce v8::Value

workerd/server/workerd-api.c++:759: error: wrapped binding module can't be resolved (internal modules only); moduleName = miniflare-internal:wrapped:__WRANGLER_EXTERNAL_AI_WORKER
workerd/jsg/util.c++:331: error: e = workerd/server/workerd-api.c++:789: failed: expected !value.IsEmpty(); global did not produce v8::Value

beelzabub•2/27/25, 11:53 PM

Ah, just looks like a straight up bug: https://github.com/cloudflare/workers-sdk/issues/7434

GitHub

🐛 BUG:Could not start vitest when using Cloudflare Vectorize, AI, a...

Which Cloudflare product(s) does this pertain to? Workers AI, Wrangler, Miniflare What version(s) of the tool(s) are you using? 3.91.0 What version of Node are you using? v22.9.0 What operating sys...

IIsaac McFadyen That's for Whisper Turbo, one of the Workers AI engineers confirmed a while back...

mr.niko.la•2/27/25, 11:57 PM

sorry misread it. i will defintately try it.

currently i am sendig audio to backend then sending to groq then to and llm provider.

this looks like a much simpler way to do it and at the very least a redudant and failover option.

right now taking 5-7s but if i am able to send earlier and depending on the edge vs groq it might be very close.

mr.niko.la•2/28/25, 1:49 AM

where is this i dont see it in the workers template no more.

mr.niko.la•2/28/25, 1:55 AM

https://youtu.be/RvJ-T5BrAzc?si=zapE1xdXpM5MJCAD&t=42
this 2min video is so much simpler than cf dev youtube channel.

YouTubeilovefreesoftwareTV

Run OpenAI’s Whisper on Cloudflare for Free using Workers AI [Speec...

#cloudflare #whisper #openai #speechtotext #machinelearning #workersai

In this video, I will show you How to run OpenAI’s Whisper on Cloudflare using Workers AI.

Before we begin, make sure you hit subscribe and tap the bell icon to get daily tech videos.

Workers AI is a new beta feature of Cloudflare that allows you to run popular AI models ...

mr.niko.la•2/28/25, 1:57 AM

althought i do appreicate their cool examples its just too complicated especially when the frontend is Astro which is another brand new framework. I do want to replicate their last teo demo but once I get this simple worker going first. UGHHHHH

mr.niko.la•2/28/25, 2:08 AM

lol cursed gui

DragoDiKomo•2/28/25, 3:55 PM

Im getting this error: 3040 : Capacity temporarily exceeded, please try again. I'm on a paid plan for AI and with an application in production

DDragoDiKomo Im getting this error: 3040 : Capacity temporarily exceeded, please try again. I...

Isaac McFadyenOP•2/28/25, 3:56 PM

This isn't a per-account limit but a global limit indicating that Workers AI as a service is above capacity.

IIsaac McFadyen This isn't a per-account limit but a global limit indicating that Workers AI as ...

DragoDiKomo•2/28/25, 3:56 PM

Image generation is working

DragoDiKomo•2/28/25, 3:56 PM

btw

Isaac McFadyenOP•2/28/25, 3:56 PM

Yeah it'll be per-model.

Isaac McFadyenOP•2/28/25, 3:56 PM

Just depending on how many replicas of that model they have deployed.

IIsaac McFadyen This isn't a per-account limit but a global limit indicating that Workers AI as ...

DragoDiKomo•2/28/25, 3:56 PM

Yup I understand but it's a problem anyway

Isaac McFadyenOP•2/28/25, 3:56 PM

For sure, just wanted to make you aware.

DragoDiKomo•2/28/25, 3:58 PM

I need to implement a fallback logic with a model that is providing at least the same quality result

fatwang2•3/1/25, 6:12 AM

hey, team. Will you think about add more long context models?

shivam•3/1/25, 8:42 AM

I'm encountering an Error: 10000: Authentication error when trying to connect to Workers AI locally. Despite successfully completing Wrangler login, it still doesn't work. However, I can connect to both local and remote D1 instances, but Workers AI remains inaccessible.

王

王誉曾•3/1/25, 9:09 AM

你好

王

王誉曾•3/1/25, 9:09 AM

王

王誉曾•3/1/25, 9:10 AM

myworker code does not

SuperHelpflare•3/1/25, 9:10 AM

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.

fatwang2•3/2/25, 4:36 AM

why the model on the dashboard is different from what I am using

James•3/2/25, 4:57 AM

Which model are you using? If it’s beta, it won’t necessarily show up in your stats/usage yet

JJames Which model are you using? If it’s beta, it won’t necessarily show up in your st...

fatwang2•3/2/25, 5:16 AM

@cf/deepseek-ai/deepseek-r1-distill-qwen-32b

fatwang2•3/2/25, 5:18 AM

this model seems the longest context model in workers AI

fatwang2•3/2/25, 5:20 AM

by the way, this model can't send reasoing content stablely

Mohamad•3/3/25, 6:53 AM

anyone here worked with dream shaper text to image model ?
i tried to get some result from it but it looks like there is a problem with generating images , take a look

Mohamad•3/3/25, 6:55 AM

to my experience they are using a wrong sampler on the server side , so cloudfalre doens't allow us to cahnge the sampler , and so this is the WRONG result we are getting from this model

Victor•3/4/25, 12:04 AM

The model page

llama-guard-3-8b

llama-guard-3-8b

input schema is just a copy paste of every other model page but it doesn't actually match what is live. For example:

```
system
```
```
system
```
and
```
tool
```
```
tool
```
are not valid message roles

tom•3/4/25, 3:14 PM

is there any way to run

pyannote

pyannote

on cloudflare?

云

云•3/5/25, 10:00 PM

云

云•3/5/25, 10:00 PM

bruh

Ttom is there any way to run `pyannote` on cloudflare?

tornado•3/5/25, 10:48 PM

It's a heavy model, I would be surprised if they add it

tornado•3/5/25, 10:50 PM

Can we get a VAD model to workers AI? The whisper model have vad so I imagine having a just VAD model can be done easily and can be very useful for voice calls.

nclevenger•3/6/25, 11:39 AM

Would love this new reasoning model released yesterday on Workers AI ... it looks like it substantially outperforms the DeepSeek R1 Qwen 32b distillation: https://huggingface.co/Qwen/QwQ-32B

Qwen/QwQ-32B · Hugging Face

Nnclevenger Would love this new reasoning model released yesterday on Workers AI ... it look...

fatwang2•3/7/25, 2:16 AM

why I can't find it on workers ai?

Ffatwang2 why I can't find it on workers ai?

Isaac McFadyenOP•3/7/25, 3:40 AM

It's not on Workers AI yet.

Isaac McFadyenOP•3/7/25, 3:40 AM

They are requesting it be added, not saying it's already on Workers AI.

MMohamad anyone here worked with dream shaper text to image model ? i tried to get some ...

Raylight•3/7/25, 9:50 AM

The only way I been able to get decent results is by setting

guidance

guidance

to something like 1, possibly 2. With higher values, problems with saturation and contrast quickly become apparent. The img2img and inpainting models had similar issues last time I checked. (See cloudflare-ai for example.) And SDXL-lightning has a somewhat similar issue. (Cartoonish speckled result unless you set

num_steps

num_steps

to e.g. 2 and

guidance

guidance

to a low value.)

lfdepombo•3/7/25, 4:51 PM

Hi, I am consistently having requests through the

ai/run

ai/run

endpoint truncated when

@cf/deepseek-ai/deepseek-r1-distill-qwen-32b

@cf/deepseek-ai/deepseek-r1-distill-qwen-32b

is clearly midsentence. The request is always 200 too. Is this a known issue of some kind? This has been happening for the past 2 days now.

lfdepombo•3/7/25, 4:52 PM

would streaming or using the OpenAI compatible API help at all?

lfdepombo•3/7/25, 5:01 PM

ah! default

max_tokens

max_tokens

for the result is only 256. bumping it up fixed it. there is no indication about it in the response though

Llfdepombo ah! default `max_tokens` for the result is only 256. bumping it up fixed it. the...

James•3/7/25, 5:07 PM

Yeah this was an undocumented breaking change. Previously the defaults weren't enforced but they have been for a short while now. We're trying to get a changelog entry or something about this so folks are aware of the change.

ajgeiss0702•3/7/25, 11:52 PM

it looks like

@cf/deepseek-ai/deepseek-r1-distill-qwen-32b

@cf/deepseek-ai/deepseek-r1-distill-qwen-32b

is missing the starting <think><think> tag in responses

workerd/server/workerd-api.c++:759: error: wrapped binding module can't be resolved (internal modules only); moduleName = miniflare-internal:wrapped:__WRANGLER_EXTERNAL_AI_WORKER workerd/jsg/util.c++:331: error: e = workerd/server/workerd-api.c++:789: failed: expected !value.IsEmpty(); global did not produce v8::Value

That's for Whisper Turbo, one of the Workers AI engineers confirmed a while back (they don't have no

Similar Threads

Similar Threads

Similar Threads