Share via

Ai Studio: Prompt flow not giving connected resources anymore

Jeffrey De Meulemeester 30 Reputation points
2024-09-27T10:21:43.3633333+00:00

Since a few days using prompt flow in AI studio (ai.azure.com) is not working anymore because there seems to be no connections available, even though the AI resources are connected to the AI Studio Project.
User's imageUser's image

We've tried:

  • reconnecting the openAI resources in the project (didn't work)
  • adding a manual connection in prompt flow through the '+ add connection' button (didn't work, we then get an error that no deployment is available.)
  • redeploying a new prompt flow from the Chat interface (didn't work)

It's occuring:

  • in all prompt flows
  • in all stepts that use a openAI connection
  • for all AI studio users

It used to work before, we've been working in prompt flows for months now, so we believe it's a recent general issue at Microsoft side. Anyone facing the same issues?

Azure OpenAI in Foundry Models

1 answer

Sort by: Most helpful
Most helpful Newest Oldest
  1. Rohit Mungi 49,131 Reputation points Microsoft Employee Moderator
    2024-09-27T12:50:24.0233333+00:00

    @Jeffrey De Meulemeester I have added couple of OpenAI connections in my AI studio and shared them to all projects and they seem to be available in the dropdown.

    User's image

    User's image

    I think you can try adding them directly in raw file mode and check if that works.

    User's image

    In raw file mode, an llm should like below along with other components that you have added. The provider should be AzureOpenAI and connection should be your connection name.

    If you are unable to use the same after trying above options, you could try reaching out to Azure support by raising a case from Azure portal. Thanks!! I hope this helps!!

    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.

    Was this answer helpful?

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.