Hello,
I am encountering a Guardrails error in Agent Builder Workflow.
Error Summary:
Guardrails Summary: Error
Inference stream error:
503 Service Unavailable
Model:
omni-mod-internal-t
Request ID:
bf6f9c33-4565-43e4-9a30-00e17f8e1f8a.ggpnxa_870
Timestamp:
2026-06-08 06:41:14 GMT
The workflow runs successfully until it reaches the Guardrails step. The error occurs during moderation/guardrail inference.
I have already tried:
Has anyone experienced this issue recently with omni-mod-internal-t?
Thank you.
We too are facing the same issue in our workflow. I found an article that suggest the moderation API in question was recently changed. It appears the moderation node in the Agent Builder graph may not have been updated yet to reflect this change. Does anyone from OpenAI monitor these threads?
Since I can’t link to the article, search for “OpenAI Folds omni-moderation-latest Into Generation Calls, Cutting App Complexity”.
If I delete the Guardrails node and recreate it, the node no longer an “Error” output connector. When I try to run the workflow, it just fails with the following message:
Workflow failed: We experienced an error while running the workflow. Sorry about that! You can retry your request, or contact us through our help center at help.openai.com if you keep seeing this error. (Please include the request ID e1b504ba-564d-438d-960c-96fe5cecb964 in your message.). (code: unknown_error)
I have gone to help.openai.com and I don’t even see a place to open a ticket.
It should be on the bottom right. Might be an ad-blocker?
Welcome to the community!
It’s the speech bubble:
When you click that, this comes to view:
I chatted with the support bot for a good 45 minutes. At first it tried to offer support for Ads Manager, despite acknowledging the issue is with Agent Builder. Eventually I got the bot back on track and it required me to send 12 different sessions, trace IDs, request IDs, etc.
The bot eventually asked me to create a new minimum workflow, which I did and the error still persists.
The bot eventually told me to turn off Guardrails. When I said that we didn’t want to lose the safeguards against self-harm, it suggested that I put in a retry loop. This retry loop repeats infinitely because the Guardrails node is never successful.
I finally said that I really needed to speak with a human, and the bot finally agreed. It said I “can expect a response in the coming days”. 
