openai/openai-agents-python

"gpt-realtime" model is not usable for the RealtimeRunner

Closed this issue · 3 comments

aligokalppeker commented

Describe the bug

The gpt-realtime model can not be used with the Realtime Agent.

Debug information

  • Agents SDK version: (v0.2.11)
  • Python version (Python 3.13)

Repro steps

When you create the runner with the "gpt-realtime" model with the following code;

RealtimeRunner(
            starting_agent=triage_agent,
            config=RealtimeRunConfig(model_settings=RealtimeSessionModelSettings(model_name="gpt-realtime")))

Then the bot fails with the Pydantic validation error.
It appears that the agent utilizes beta versions of the OpenAI SDK models.

Expected behavior

Expecting the model to be used with the general available model. The same model can be used when connecting to the "realtime" agent through the raw WebSocket interface.

seratch commented

Thanks for writing in. We're still working on the gpt-realtime support at #1646; once the PR is merged and we release a new version, you'll be able to use gpt-realtime with this SDK.

seratch commented

Please check #1614 for progresses instead

  • 👍1
aligokalppeker commented

Thanks for your efforts @seratch