I'm currently working on implementing AI services within our organization, and we've set up private connectivity as part of our cybersecurity measures. Foundry is now configured to run exclusively on private endpoints that we can connect to via express routes. While this setup is functional, it completely restricts users from accessing the inference (chat playground) features of Foundry, which means we're limited to API access only. Is this just how it operates, or is there a workaround to enable playground access? I noticed that the ai.azure.com site is external to our tenant and primarily makes API calls, which seems to block access to our custom Foundry deployment when we disable public access. How are others in the community managing this issue? Similar situations have been noted with older versions of Foundry and even back when only Azure OpenAI was available.
3 Answers
Any chance you have links to reference articles on this? I saw someone else mention that it actually works in some cases, so I’m curious if there are any official statements out there.
What specific Foundry resource are you using right now? The newer one based on Cognitive Services, or are you still on the classic hub experience? My experience has been mostly with the classic version linked to Azure Machine Learning workspaces.
The latest I heard is that this problem is known and should be resolved in the near future. It’s a bit frustrating, but hopefully, there will be a fix soon.
Related Questions
Biggest Problem With Suno AI Audio
How to Build a Custom GPT Journalist That Posts Directly to WordPress