Unanswered

Azure AI Foundry Inference connector?

(1) Share

Report

Posted on by chrisadriane

Does anyone have experience with this connector? When trying to use it i get "Error Message: The connector 'Azure AI Foundry Inference' returned an HTTP error with code 404. Error Code: ConnectorRequestFailure". When i connect to the connector it looks like the image. I am not sure if this is how it is supposed to look like as it doesn't look like it really is connected?

For the connection can someone clarify on which Uri should be set here? As i in my Foundry Playground have tried all of the endpoints to the project (image below) as well as the dicerct Uri to my concrete model (for example the gpt-4o). And none of them seem to work.

Categories:

General topics

I have the same question (3)

All responses (7)

Answers (0)

Romain The Low-Code... 2,111 Super User 2025 Season 2 on at

Like (0)

Report

to folow too, since i prefere to use the "deploy to copilot studio" model from ai foundry playground, i don't play with this one, so i m curious :)

Was this reply helpful? Yes No
chrisadriane 23 on at

Like (0)

Report

@Romain The Low-Code Bearded Bear

It seems I don't have that option. When i Open my model in playground it looks like i can only Deploy as a web app. Do you experience the same issue or does it allow you to deploy to Copilot Studio?

Was this reply helpful? Yes No
Romain The Low-Code... 2,111 Super User 2025 Season 2 on at

Like (0)

Report

@chrisadriane many things could have change this the MsBuild monday but i made a demos on may the 7th and i had to change Change the url from Azure AI Foundry to https://oai.azure.com/

since it was in preview. Probably manythings have changed since the MsBuild.

But it s a different things , inference are not the same , i was juste speaking about this case because i never try your way :) so i m curious on the answer :)

Was this reply helpful? Yes No
Suggested answer

MM-18061923-0 4 on at

Like (1)

Report

I get it work by fidling with endpoint URL. It should end on GPT, Like this:

https://xyz.cognitiveservices.azure.com/openai/deployments/gpt-4.1

Where xyz is your name of Azure AI Foundry instance. It it in default url.

And use api version

1 people found this reply helpful.

Was this reply helpful? Yes No
WCarterMVUSD 87 on at

Like (0)

Report

I'm getting a similar issue, whenever I try to use "Create a Chat Completion" I get

"error": {

            "code": "404",

            "message": "Resource not found"

        }

I'm not sure what to do with this. I've tried many permutations of the target uri, of parameters like model name, nothing seems to work. Any help would be appreciated!

Was this reply helpful? Yes No
Suggested answer

RMIC 2 on at

Like (0)

Report

This worked for me.

Go to your model in Azure AI Foundry and copy the URL and Key of your model:

The URL must be modified to be accepted:

https://<xyz>.cognitiveservices.azure.com/openai/deployments/gpt-4o/chat/completions?api-version=2025-01-01-preview

You need to delete everything after your model name:

https://<xyz>.cognitiveservices.azure.com/openai/deployments/gpt-4o

With this information you can add a new Connections "Azure AI Foundry Inference":

Display Name: <name of your connections>

Target URI: https://<xyz>.cognitiveservices.azure.com/openai/deployments/gpt-4o

API Key: Key from your model (see above)

Model deployment name: Name of your model. I used gpt-4o

Base Model name: I used the same information as before: gpt-40

Next, add in Power Automate the action "Create a chat completion":

API version: 2025-01-01-preview (see also the URL above after "api-version")

Body/Message:

[

{

"role": "system",

"content": [

{

"type": "text",

"text": "You are a helpful assistant."

}

]

},

{

"role": "assistant",

"content": [

{

"type": "text",

"text": "I am going to Paris, what should I do?"

}

]

}

]

Body/temperature: I tested with: 0.1

Body/top: I tested with: 0.1

Body/max tokens: I tested with: 100

Body/model: gpt-40

Was this reply helpful? Yes No
CW-07110951-0 3 on at

Like (0)

Report

I encountered this issue too.

Based on @MM-18061923-0 's answer, here is my final configuration. It should work if you follow this screenshot.

Good luck.

Was this reply helpful? Yes No