I don't have one because I don't know the answer to their question, but telling them to use Appscrip

IIsaac McFadyen I don't have one because I don't know the answer to their question, but telling ...

C

Carlos Alberto•7/11/25, 8:15 PM

Why not?

CCarlos Alberto Why not?

I

Isaac McFadyenOP•7/11/25, 8:17 PM

Because... it's completely unrelated? It's not the Cloudflare product they asked about and it's not the answer to their (very specific) question.

IIsaac McFadyen Because... it's completely unrelated? It's not the Cloudflare product they asked...

C

Carlos Alberto•7/11/25, 8:18 PM

Is that a fact?

Ddan—1106 for folks who come across this later, I attached files to my messages like this:...

V

Victor•7/11/25, 8:19 PM

Just FYI, there's an attachment limit of 20MB with this format as documented in the Files API. Beyond that (or as an alternative to reuse the same upload for multiple prompts), they recommend using their files API and using the reference (from the returned metadata) to it as the new

image_url

image_url

in the prompt. Full File API Endpoint documentation

V

Vk•7/26/25, 11:15 PM

Question on this https://developers.cloudflare.com/ai-gateway/configuration/authentication/ > Does this create a token scoped to the specific gateway only ? and can we generate key programmatically?

Cloudflare Docs

Authentication

Add security by requiring a valid authorization token for each request.

VVk Question on this https://developers.cloudflare.com/ai-gateway/configuration/auth...

V

Vk•7/28/25, 6:10 PM

Other issue i had was even from my account, when i tried to create a token...it was not working...

T

Tristan•7/30/25, 9:59 AM

Hello, we've been using GatewayAI since almost day one, but it's still sad to see that you don't put prices for

openAI/dall-e-3

openAI/dall-e-3

image generation.

Our apps uses a lot of that and the pricing you're displaying is wrong since you don't take that into consideration.

One easy fix and accurate would be to map prices with the quality and resolution asked since it's known there:

S

Shados•8/1/25, 6:30 PM

Does CF AI Gateway support enteprise log push? Reading the history in this channel and the few convos I can see are unclear (not the documented Worker Log Push). We use it for all our Worker stuff without issue, but having trouble getting it to work with CF AI Gateway.

S

Shados•8/1/25, 6:44 PM

Its clearly there, but no data goes through for us and it doesn't seem to be anywhere in the documentation, as if its not supposed to exist.

W

wellroundedtable•8/1/25, 6:56 PM

Hey guys, I think I may have complicated the system by trying to manage my domain from multiple Cloudflare acounts. In one account my domains status is: active. Cloudflare sent an email to that email address warning me that I need to validate my domain, but when I click the box and approve the certificate I receive this error:

An error occurred while attempting to validate your domain. Please try again later or contact support for assistance.

Does this have anything to do with my being “connected” to the same domain with a different Cloudflare acounts associated with a different email address with the status as “invalid nameservers.”

B

Bash•8/5/25, 2:54 PM

AI gateway doesn't support transcription endpoints? I tried sending the below request through the gateway but it just gives this error: "500 Audio file processing failed". Looks like something to do with the audio file not getting sent properly? It works fine when sending it without AI Gateway

Request:

{
  "model": "gpt-4o-mini-transcribe",
  "response_format": "text",
  "file": {}
}

{
  "model": "gpt-4o-mini-transcribe",
  "response_format": "text",
  "file": {}
}

{
  "model": "gpt-4o-mini-transcribe",
  "response_format": "text",
  "file": {}
}

{
  "model": "gpt-4o-mini-transcribe",
  "response_format": "text",
  "file": {}
}

Response:

{
  "error": {
    "message": "Audio file processing failed",
    "type": "server_error",
    "param": null,
    "code": null
  }
}

{
  "error": {
    "message": "Audio file processing failed",
    "type": "server_error",
    "param": null,
    "code": null
  }
}

{
  "error": {
    "message": "Audio file processing failed",
    "type": "server_error",
    "param": null,
    "code": null
  }
}

{
  "error": {
    "message": "Audio file processing failed",
    "type": "server_error",
    "param": null,
    "code": null
  }
}

S

Sam•8/6/25, 5:49 AM

Is it possible to help me with getting cloudflare AI gateway + azure + vercel ai sdk running? (vercel ai sdk uses the /openai/v1/chat/completions approach rather than the deployment name like the azure provider cloudflare ai gateway uses.

S

Sam•8/6/25, 5:50 AM

Have tried a bunch of combinations and not managed to get anything working! (it works fine with all my other providers)

S

saeedesmaili•8/7/25, 11:36 AM

Is it possible to set

thinking_budget

thinking_budget

when using

google-ai-studio

google-ai-studio

models on ai gateway? Google has thinking enabled by default for models like

gemini-flash-2.5

gemini-flash-2.5

, and to request a response without thinking we should set

thinking_budget

thinking_budget

to 0, but I can't figure out how to do this when using cloudflare ai gateway.

V

vcd•8/12/25, 8:51 AM

Hello everyone, I encountered some issues while using ai-gateway.

I usually use OPTIONS requests to check if the interface is available, but when I request openrouter through the gateway, I get a 500 error internally in Cloudflare, like this

curl --location --request OPTIONS ‘https://your-gateway/openrouter/v1/models’

curl --location --request OPTIONS ‘https://your-gateway/openrouter/v1/models’

which would return

< HTTP/2 500
< date: Tue, 12 Aug 2025 08:49:47 GMT
< content-type: application/json
< content-length: 101
< server: cloudflare
< nel: {"report_to":"cf-nel","success_fraction":0.0,"max_age":604800}
< vary: Accept-Encoding
< cf-cache-status: DYNAMIC
< strict-transport-security: max-age=15552000; includeSubDomains; preload
< x-content-type-options: nosniff
< cf-ray: 96deb8af99b71d6d-SIN
< alt-svc: h3=":443"; ma=86400

{"success":false,"result":[],"messages":[],"error":[{"code":2002,"message":"Internal server error"}]}

< HTTP/2 500
< date: Tue, 12 Aug 2025 08:49:47 GMT
< content-type: application/json
< content-length: 101
< server: cloudflare
< nel: {"report_to":"cf-nel","success_fraction":0.0,"max_age":604800}
< vary: Accept-Encoding
< cf-cache-status: DYNAMIC
< strict-transport-security: max-age=15552000; includeSubDomains; preload
< x-content-type-options: nosniff
< cf-ray: 96deb8af99b71d6d-SIN
< alt-svc: h3=":443"; ma=86400

{"success":false,"result":[],"messages":[],"error":[{"code":2002,"message":"Internal server error"}]}

< HTTP/2 500
< date: Tue, 12 Aug 2025 08:49:47 GMT
< content-type: application/json
< content-length: 101
< server: cloudflare
< nel: {"report_to":"cf-nel","success_fraction":0.0,"max_age":604800}
< vary: Accept-Encoding
< cf-cache-status: DYNAMIC
< strict-transport-security: max-age=15552000; includeSubDomains; preload
< x-content-type-options: nosniff
< cf-ray: 96deb8af99b71d6d-SIN
< alt-svc: h3=":443"; ma=86400

{"success":false,"result":[],"messages":[],"error":[{"code":2002,"message":"Internal server error"}]}

< HTTP/2 500
< date: Tue, 12 Aug 2025 08:49:47 GMT
< content-type: application/json
< content-length: 101
< server: cloudflare
< nel: {"report_to":"cf-nel","success_fraction":0.0,"max_age":604800}
< vary: Accept-Encoding
< cf-cache-status: DYNAMIC
< strict-transport-security: max-age=15552000; includeSubDomains; preload
< x-content-type-options: nosniff
< cf-ray: 96deb8af99b71d6d-SIN
< alt-svc: h3=":443"; ma=86400

{"success":false,"result":[],"messages":[],"error":[{"code":2002,"message":"Internal server error"}]}

I'm not sure why this is happening, as other endpoints are functioning normally.

Original message was deleted

I

Isaac McFadyenOP•8/14/25, 3:13 AM

?crossposting also this is a better fit for Workers AI, the NSFW detector isn't related to AI Gateway.

S

SuperHelpflare•8/14/25, 3:13 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.

IIsaac McFadyen ?crossposting also this is a better fit for Workers AI, the NSFW detector isn't ...

N

Nhat Bui•8/14/25, 3:15 AM

thanks, I am deleting

V

Victor•8/14/25, 7:06 PM

Trying to check logs for an ai gateway and getting HTTP 500 (

)

Error: The requested Durable Object is temporarily disabled, please contact us for details

Error: The requested Durable Object is temporarily disabled, please contact us for details

V

Victor•8/15/25, 6:31 PM

I'm getting 404

7002: Not Found

7002: Not Found

on

Patch Gateway Log

Patch Gateway Log

. But going into the ai gateway dashboard and I can find the exact log from the url

M

mrwind•8/20/25, 1:54 AM

hi can i use AI gateway but route request from hongkong to another regions ? I got this error "message": "User location is not supported for the API use.", "status": "FAILED_PRECONDITION"

J

John Reyes•8/20/25, 3:03 AM

https://athleteoftheyear.org/2025/luke-72b9

Vote for Luke to be featured in a 3BRAND ad in Sports Illustrated a...

8 Year Old Amazing Baseball/soccer Player Who Works Hard Day In And Day Out. Started At Age 4 And Currently Plays Up In 9u And 11u Teams.

J

John Reyes•8/20/25, 3:04 AM

Vote for this kid

Mmrwind hi can i use AI gateway but route request from hongkong to another regions ? ...

V

vcd•8/20/25, 6:57 AM

AIGateway will default to inserting Cf-Country into the request, which cannot be changed.

When you request a provider such as OpenAI, it will be filtered according to the provider's service policy.

For similar providers, I recommend using a gateway service such as OpenRouter or adding an additional layer of routing (to SG maybe).

K

Kevin•8/20/25, 10:00 AM

In case anyone is wondering since its not written anywhere, gateway does support responses enpoint using workers-ai provider for gpt-oss models. I was just testing that

K

Kevin•8/20/25, 5:01 PM

The cost feature is automatic or do I have to send it? Because the cost is still 0 while using workers-ai provider.

Y

yinxingmaiming6409•8/21/25, 2:11 AM

How can AI be used in Cherry Studio? It seems impossible.

Yyinxingmaiming6409 How can AI be used in Cherry Studio? It seems impossible.

I

Isaac McFadyenOP•8/21/25, 2:37 AM

?crossposting Please don't post in multiple channels.

S

SuperHelpflare•8/21/25, 2:37 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.

T

Tristan•8/21/25, 7:46 AM

hey, can someone from cloudflare can take a look at my message, it slipped through: ai-gateway

B

ben_makes_stuff•8/25/25, 2:57 AM

Any plans to make the OpenAI-compatible endpoint work with Amazon Bedrock? They're very lazy about their implementation and recommend customers use an AWS lambda function with custom code if using the OpenAI SDK, they don't provide an OpenAI-compatible endpoint like Google does

D

dave•8/26/25, 6:51 PM

Looks like pricing for Anthropic models with caching has been fixed?

H

hurrytospring•8/28/25, 5:34 AM

We are a startup tech company considering using Cloudflare AI Gateway. Before proceeding, we have the following questions:
It seems that there is no support for the DeepSeek V3 or Qwen 3 models. Is there any other way to access them? They are really cost - effective.
Is the AI Gateway ready for production use? What are the limits of queries per second (QPS) and tokens per minute (TPM)?
Compared to native providers like Vertex/ Bedrock, when making requests from Google Cloud Platform (GCP) or Amazon Web Services (AWS), how does the latency or token speed differ? By the way, how does the price compare?

W

wangzuo•8/28/25, 2:08 PM

Can we specify the AI gateway region? Some Cloudflare proxy servers in certain locations do not have access to specific AI providers.

A

ari•8/29/25, 3:17 AM

AI Gateway behavior appears to have changed this afternoon, and requests that were previously returning successfully after 3 minutes are now failing after 60 seconds. (I'm using AI Gateway for requests to a number of different vendor models, including Anthropic's Opus 4.1, where responses to complex requests can take 1-3 minutes to generate.) Cloudflare is reporting this as a 504, but when I looked at the corresponding log in Anthropic, they're reporting it as a 499 "client disconnected" after 60 seconds, suggesting that Cloudflare terminated the connection. This definitely breaks slow models like Opus. Has anyone else run into this today?

D

dave•8/29/25, 3:54 PM

https://developers.cloudflare.com/ai-gateway/configuration/bring-your-own-keys/

Great job!!

Cloudflare Docs

BYOK (Store Keys)

Bring your own keys (BYOK) is a feature in Cloudflare AI Gateway that allows you to securely store your AI provider API keys directly in the Cloudflare dashboard. Instead of including API keys in every request to your AI models, you can configure them once in the dashboard, and reference them in your gateway configuration.

Aari AI Gateway behavior appears to have changed this afternoon, and requests that we...

A

ari•8/29/25, 11:04 PM

(Seems like this was a temporary issue. Longer calls are working again.)

Aari AI Gateway behavior appears to have changed this afternoon, and requests that we...

D

dave•8/30/25, 1:09 AM

Was this for streaming requests?

W

Ward•8/30/25, 8:33 PM

Hey all, is there an easy way to store my keys in the cloudflare dashboard and then call the unified api? When i do this I still have to provide the api key of my underlying provider. How did you fix this? Here is an example with cerebras

const client = new OpenAI({
            apiKey: apiKey,
            baseURL: 'https://gateway.ai.cloudflare.com/v1/{account_id}/{gateway_id}/compat',
            defaultHeaders: {
                'cf-aig-authorization': `Bearer ${apiKey}`,
            },
        });

        const response = await client.chat.completions.create({
            model: 'cerebras/gpt-oss-120b',
            messages: [{ role: 'user', content: 'What is cloudflare?' }],
        });

const client = new OpenAI({
            apiKey: apiKey,
            baseURL: 'https://gateway.ai.cloudflare.com/v1/{account_id}/{gateway_id}/compat',
            defaultHeaders: {
                'cf-aig-authorization': `Bearer ${apiKey}`,
            },
        });

        const response = await client.chat.completions.create({
            model: 'cerebras/gpt-oss-120b',
            messages: [{ role: 'user', content: 'What is cloudflare?' }],
        });

const client = new OpenAI({
            apiKey: apiKey,
            baseURL: 'https://gateway.ai.cloudflare.com/v1/{account_id}/{gateway_id}/compat',
            defaultHeaders: {
                'cf-aig-authorization': `Bearer ${apiKey}`,
            },
        });

        const response = await client.chat.completions.create({
            model: 'cerebras/gpt-oss-120b',
            messages: [{ role: 'user', content: 'What is cloudflare?' }],
        });

const client = new OpenAI({
            apiKey: apiKey,
            baseURL: 'https://gateway.ai.cloudflare.com/v1/{account_id}/{gateway_id}/compat',
            defaultHeaders: {
                'cf-aig-authorization': `Bearer ${apiKey}`,
            },
        });

        const response = await client.chat.completions.create({
            model: 'cerebras/gpt-oss-120b',
            messages: [{ role: 'user', content: 'What is cloudflare?' }],
        });

apiKey is the gateway key.
Or did I get it wrong that the unified api (openai compatible) doesn't yet work with the BYOK (Store keys) feature?
Would love to hear if anyone knows.

Ddave Was this for streaming requests?

A

ari•9/1/25, 12:05 AM

No. Just a basic call to

/anthropic/v1/messages

/anthropic/v1/messages

/anthropic/v1/messages

/anthropic/v1/messages

J

joe•9/1/25, 12:26 AM

am I misunderstanding how metadata filtering is supposed to work? if I attach say

asn

asn

asn

asn and

ip

ip

ip

ip to the

metadata

metadata

metadata

metadata key I cannot actually negatively filter them

J

joe•9/1/25, 12:27 AM

??

J

joe•9/1/25, 12:27 AM

i don't know which role is for moderators here but spamming in a load of channels isn't a good look

J

joe•9/1/25, 12:28 AM

if I do

Metadata Value

Metadata Value

Metadata Value

Metadata Value <does not equal>

Some IP

Some IP

Some IP

Some IP then it negatively matches all the ASNs because they do not have the value of the IP

I don't have one because I don't know the answer to their question, but telling them to use Appscrip

Similar Threads

Similar Threads

Similar Threads