How to create embeddings client from Azure AI Foundry project's connected embedding model via Python Azure SDK

Question

How to create embeddings client from Azure AI Foundry project's connected embedding model via Python Azure SDK

Paris Perlegkas 80

I am normally connecting to my Azure AI Foundry project in which I have deployed and connected an embedding model (pls check below python script attached image). When I am trying to create a vector embeddings client with project.inference.get_embeddings_client()via python Azure SDK, I get error "No connection of type ConnectionType.AZURE_AI_SERVICES found", and as a result no embedding is found, hence no document search can be performed later on - I am building a RAG application.

What am I missing? I don't see any ConnectionType.AZURE_AI_SERVICES in documentation https://learn.microsoft.com/en-us/python/api/azure-ai-projects/azure.ai.projects.models.connectiontype?view=azure-python-preview.

Any help would be highly appreciated. Thanks in advance.

self.project = AIProjectClient.from_connection_string(
conn_str=os.environ["AIPROJECT_CONNECTION_STRING"], 
credential=DefaultAzureCredential()
)

            # Create a project client using environment variables
            self.logger.info(f"Initializing AI Project client with connection string")
            self.project = AIProjectClient.from_connection_string(
                conn_str=os.environ["AIPROJECT_CONNECTION_STRING"], 
                credential=DefaultAzureCredential()
            )

            # Try to create vector embeddings client
            try:
                self.embeddings = self.project.inference.get_embeddings_client()
            except ResourceNotFoundError as e:
                self.logger.warning(f"ResourceNotFoundError: No embeddings service found in project: {str(e)}")

User's image

1 answer

Your answer

Answer 1

Ravada Shivaprasad 1,965 Microsoft External Staff Moderator

Hi Paris Perlegkas

The error appears to stem from a missing or misconfigured connection to Azure AI Services within your Azure AI Foundry project. The system does not recognize ConnectionType.AZURE_AI_SERVICES, which suggests that this connection type may not be available in the Azure AI Projects SDK.

To resolve this, first, review the available connection types in the Azure AI Projects SDK documentation to confirm whether an alternative type, such as ConnectionType.AI_SERVICES, should be used. Additionally, verify that your connection string stored in the environment variable AIPROJECT_CONNECTION_STRING is correctly set and pointing to the appropriate Azure AI Foundry project.

If the issue still persists, ensure that your project has an active connection to Azure AI Services by navigating to Management Center → Connections → New Connection in the Azure AI Foundry Portal. Proper error handling is also crucial—your ResourceNotFoundError exception should provide more insights into why the embeddings service is not found.

Lastly, check for any SDK updates or changes that might have impacted connection types or embedding client creation.

I Hope this helps. Let me know if you need further clarification!

If this answers your query, please do click Accept Answer and Yes for was this answer helpful.

Thank you!

Paris Perlegkas 80 Reputation points

2025-05-29T08:40:35.58+00:00

I have already reviewed all available connection types, and I already know that the connection type that the error is suggesting (ConnectionType.AZURE_AI_SERVICES) does not exist in Azure AI Projects SDK, which is why the displayed error confuses and complicates more rather than helps. You can also see from my code that I don't use the connection type, therefore that's not the reason it doesn't work.

Furthermore, by reviewing this step by step tutorial, both my project connection and embedding client setup look correct. And I can confirm that the project connection works correctly, and it is normally reading the AZURE_OPEN_AI connection that I have set up in it, and in which there is the embedding model (as also displayed in my previous attached picture).

Please let me know if you have any other idea - thanks.
Deleted

This comment has been deleted due to a violation of our Code of Conduct. The comment was manually reported or identified through automated detection before action was taken. Please refer to our Code of Conduct for more information.
Ravada Shivaprasad 1,965 Reputation points Microsoft External Staff Moderator

2025-05-31T01:10:20.4733333+00:00
Hi Paris Perlegkas

Thank you for your update. Based on your code and the error message you're encountering, it's clear that the issue is not with your implementation but rather with how the Azure AI Projects SDK is handling or reporting the error internally.
The error message referencing ConnectionType.AZURE_AI_SERVICES is misleading because this connection type does not exist in the Azure AI Projects SDK. According to the official SDK documentation, the valid connection types include AzureOpenAI, AzureAISearch, AzureBlobStorage, CosmosDB, and others—but not Azure_AI_SERVICES

. Since your code does not reference this connection type and you're using a valid one (AZURE_OPEN_AI), the error message likely stems from an internal fallback or misconfiguration in the SDK.

Your code for initializing the AIProjectClient and attempting to create the embeddings client is correct:

self.project = AIProjectClient.from_connection_string( conn_str=os.environ["AIPROJECT_CONNECTION_STRING"], credential=DefaultAzureCredential() ) self.embeddings = self.project.inference.get_embeddings_client()

The ResourceNotFoundError you're encountering likely means the SDK cannot locate an embeddings-capable model in your Azure AI Foundry project. This is not due to your code or connection type—your use of AZURE_OPEN_AI is correct. The error message referencing ConnectionType.AZURE_AI_SERVICES is misleading, as this connection type does not exist in the Azure AI Projects SDK.

This issue may occur if No embedding model (e.g., text-embedding-ada-002) is deployed under your AZURE_OPEN_AI connection or else The model is deployed but not recognized as an embeddings model or if there's a mismatch in region or configuration between the model and the project.
To Resolve this:
Even though you has confirmed that the connection itself is active and correctly configured, the reason I still recommend verifying the embedding model under that connection is because the get_embeddings_client() method doesn't just check for a valid connection—it specifically looks for a deployed and recognized embeddings-capable model within that connection.
And also ensure whether you are using the latest SDK version or not by (pip install --upgrade azure-ai-projects) meanwhile enable debug logging to trace SDK behavior and try a minimal test script to isolate the issue.

Reference : Refer to the official SDK documentation for connection types and the Python client overview.

Thanks
Ravada Shivaprasad 1,965 Reputation points Microsoft External Staff Moderator

2025-06-03T16:58:52.11+00:00

Hi Paris Perlegkas

Just check if the answer above helped you. If it answers your question, click and Accept Answer Yesfor this answer was helpful to you. And, if you have any other questions, let us know.

THANKS
Ravada Shivaprasad 1,965 Reputation points Microsoft External Staff Moderator

2025-06-05T21:12:04.1133333+00:00

Hi Paris Perlegkas

Following up to see if the above answer was helpful. If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Thanks

Share via

How to create embeddings client from Azure AI Foundry project's connected embedding model via Python Azure SDK

1 answer

Your answer