(Open ai realtime) The server had an error while processing your request. Sorry about that!

API
1

I’ve been running into this error a lot using gpt-realtime through pipecat package, is there a cause or reason for it to happen?

A fatal error occurred: ErrorFrame#11(error: The server had an error while processing your request. Sorry about that! Please contact us through our help center at ``help.openai.com`` if the error persists. (include session ID in your message: sess_CpXiXDvntsA7PRZOVS55H). We recommend you retry your request., fatal: True)

2

Sorry about that @Saif_Kharouf :frowning:

I will able to find your session. This is from an assert on our side. I believe it’s because of transcriptions.

I will work on getting more info today!

1 Like
3

This happens right after we get a conversation.item.truncate (takes it down to 0)

Then right after I see AssertionError: New text length 0 is less than last length 726. I think we have some interaction where Truncation is breaking things.

Is it easy to disable sending those truncates in pipecat and see if that makes the bug go away?

4

I’ve also suddenly started getting the same error when this was working previuosly. Here’s the log:

2026-01-19 15:22:31,805 ERROR OpenAI Response Failed: {‘type’: ‘failed’, ‘error’: {‘type’: ‘server_error’, ‘code’: None, ‘message’: ‘The server had an error while processing your request. Sorry about that! Please contact us through our help center at help.openai.com if the error persists. (include session ID in your message: sess_Czq3XSCTk82H4fjerbryP). We recommend you retry your request.’}}

And my parameters

{
  "type": "realtime",
  "model": "gpt-realtime",
  "output_modalities": ["audio"],
  "audio": {
    "input": {
      "format": { "type": "audio/pcmu" },
      "turn_detection": { "type": "semantic_vad" },
      "transcription": {
        "model": "gpt-4o-transcribe",
        "prompt": "",
        "language": "en"
      },
    },
    "output": {
      "format": { "type": "audio/pcmu" },
      "voice": "cedar",
    },
  },
  "instructions": ...,
}
1 Like
5

Hi and welcome to the community!

This is a topic from last month.
My realtime app is working fine right now but will monitor for more reports.

1 Like
6

It looks like it might have to do with the number of options I provide within the tools.allowed_tools.

1 Like
7

Limiting the number of allowed_tools options seems to have resolved that bug, but I am now getting a mcp_list_tools.failed with no further logs or details..

{
    "type": "mcp_list_tools.failed",
    "event_id": "event_D0VZ0dNHfywrWZxnzVTU6",
    "item_id": "item_D0VYzKIG80CGEE9lEKxGG"
}