Agent taking too long to return answers

[esandoval30]

I built an agent 2 weeks ago. The agent has 4 python tools and one MCP server.
It's kyc_agent.py here

Today, the same agent with the same tools and the same sequence of questions is taking over one minute and sometimes up to two minutes to return an answer.

I can see in traces that the tools are executing fast (1-2 seconds at the most) but I can't tell what is taking so long to produce the final answer that is sent by the agent.

No code changes in the last 2 weeks.

I can provide traces showing the huge difference in latency.

I tried changing default model (gpt-4o) to o3-mini, o4-mini and same result

What would you suggest to troubleshoot this problem?

[rm-openai]

Is it the actual model call the thing that is taking long? A request ID would be especially helpful

[esandoval30]

I can't really tell from Traces...but this is what I can see

SpanID = span_ebbe811d070a43c0a9688f3a
Within this SpanID, there are several items. The ones that take the longest are

POST /v1/responses takes 88 seconds
resp_685e7e579ec4819a82ba17e2fe8aa46d03ddb99fc675e75f

POST /v1/responses takes 25 seconds
resp_685e7eaeab28819aaa308392c840698d03ddb99fc675e75f

[rm-openai]

@esandoval30 we had some elevated latencies earlier today: https://status.openai.com/incidents/01JYS4X52EAB39YXTGJEYY6YQH

Things should be better now - lmk if not?

[esandoval30]

Thanks @rm-openai ! my agent is back to normal now that the elevated latencies have been resolved. Closing issue