VAPI•2y ago

Azure model provider key issue

Hello I'm having trouble adding my Azure OpenAI provider keys. I've tried two different deployments in different regions but can't get passed a "Couldn't Validate Azure credentials: Resource Not Found" error.
Is this the correct format for the endpoint URL? https://{resource}.openai.azure.com/openai/deployments/{deployment}

Ask Inkeep•6/4/24, 5:46 PM

Hi @Tripp (Summit), I'm an AI assistant for Vapi. I'll analyze available sources and get back to you shortly.

Ask Inkeep•6/4/24, 5:46 PM

-----------------------------
While I generate an answer, here are some sources that might be helpful:

Call Ended Reason
Provider Keys
Web Calling
Create Credential
-----------------------------

Ask Inkeep•6/4/24, 5:46 PM

Answer

Hello! It seems like you might be facing an issue with the format of your Azure endpoint URL or the way your API keys are being validated. While I don't have specific details on the correct URL format for Azure provider keys in the context of Vapi, typically the endpoint URL should match the format provided by Azure's documentation for accessing specific resources.

To troubleshoot your issue:

Ensure that the {resource}{resource} and {deployment}{deployment} placeholders in the URL are replaced with the actual resource name and deployment identifier provided by Azure.
Verify that your Azure API keys have the necessary permissions and are correctly configured for the resources you are trying to access.

If you continue to experience issues, it might be helpful to check the configuration settings in the Azure portal and ensure that all details are correctly entered and that there are no restrictions that might be blocking your requests.

For more detailed assistance, consulting Azure's official documentation or reaching out to their support might provide you with the specific guidance needed for your setup.

living-lavender•6/4/24, 5:48 PM

Can you first check from your end whether your API key is correct? By making an request to your endpoint with your key?

military-pinkOP•6/4/24, 5:55 PM

Hey @Sahil Ive confirmed the api key is correct, as I successfully curl the /chat/completions endpoint with the api key

living-lavender•6/5/24, 3:39 AM

can you send me any call_id or anything?

military-pinkOP•6/5/24, 5:43 PM

@Sahil unfortunately i cant start a call yet because im having trouble with the initial integration

military-pinkOP•6/5/24, 6:13 PM

Here's an example curl using the same credentials which succeeds

Mmilitary-pink Here's an example curl using the same credentials which succeeds

verbal-lime•6/5/24, 8:45 PM

Following. Tried adding azure keys today and ran into the same issues

living-lavender•6/5/24, 9:06 PM

Will look into it by tomorrow.

verbal-lime•6/6/24, 11:30 AM

@Tripp (Summit) @Sahil I managed to add the credential.

region: east - wasn't working for me, I kept getting errors saying wrong API keys or endpoint.

I had to use east 2, and name the deployment "gpt-4o-2024-05-13" Exactly how vapi expects it

Vverbal-lime @Tripp (Summit) @Sahil I managed to add the credential. region: east - wasn't w...

living-lavender•6/6/24, 11:35 AM

Thanks for the update!

military-pinkOP•6/6/24, 1:29 PM

@Jotham thanks that worked for me as well. @Sahil one idea is to add a "deployment name" field to the azure openai provider section since I doubt most people will typically name it what Vapi currently expects. Alternatively, maybe some help text explaining what the deployment name should be would be helpful

Thanks both of you!!

military-pinkOP•6/6/24, 1:30 PM

Oh and fwiw i didnt try any other regions but us-west-3 worked for me

living-lavender•6/13/24, 6:28 AM

@Tripp (Summit) @Jotham Which doc link did you guys followed?

Lliving-lavender @Tripp (Summit) @Jotham Which doc link did you guys followed?

verbal-lime•6/13/24, 1:35 PM

https://docs.vapi.ai/assistants/custom-llm

Vapi

Custom LLM - Vapi

Use Another LLM or Your Own Server

Vverbal-lime https://docs.vapi.ai/assistants/custom-llm

living-lavender•6/13/24, 3:51 PM

I mean to fix this credential issue.

Lliving-lavender I mean to fix this credential issue.

verbal-lime•6/13/24, 4:33 PM

Oh I just found the fix from trial and error. You could maybe change the name of the "model" field to "deployment" then in the info part write "azure deployment name must match exactly how Vapi expects eg: azure deployment name: gpt-4o-2024-05-13"

living-lavender•6/13/24, 4:35 PM

Ah, I see. Okay!

military-pinkOP•6/17/24, 6:30 PM

@Jotham curious are you being charged for your azure llm usage even though your provider keys are entered and being used?

Mmilitary-pink @Jotham curious are you being charged for your azure llm usage even though your...

verbal-lime•6/17/24, 6:36 PM

Yes! Forgot to question that when I saw it. thanks for bringing it up. Is that a bug @Sahil

verbal-lime•6/17/24, 6:41 PM

@Tripp (Summit) I assume you can also see your azure usage being used with every call.

Have you found where it shows your azure usage costs within the azure platform?

I can't find mine - maybe Vapi handles the azure billing side, but uses our usage?

I'm abit lost in the azure platform

Vverbal-lime @Tripp (Summit) I assume you can also see your azure usage being used with every...

military-pinkOP•6/17/24, 6:44 PM

I have not been able to find a good cost graph in azure either but this is my first experience with azure... the best Ive found is this monitoring workbook and manually calculating cost using the token charts: https://techcommunity.microsoft.com/t5/fasttrack-for-azure/azure-openai-insights-monitoring-ai-with-confidence/ba-p/4026850

Azure model provider key issue

Answer

Similar Threads

Similar Threads

Answer

Similar Threads