Realtime agent goes silent after tool calls: set initial_model_settings on .run()

TL;DR

You have to set initialmodelsettings= on the.run() call itself too. If your realtime agent calls tools but goes silent afterwards and the trace is empty, check that your model settings are set in both places, not just on the runner.

Steps

1. Reporter-verified: settings passed to RealtimeRunner are not forwarded when you call .run(). You have to set initialmodelsettings= on the .run() call itself too. Once the reporter duplicated the model settings onto .run(), the tool results flowed back into the model and tracing showed the tool calls. If your realtime agent calls tools but goes silent afterwards and the trace is empty, check that your model settings are set in both places, not just on the runner.

When to use

You are seeing this: Issue openai/openai-agents-python#1156 (closed, 37 comments): In a RealtimeAgent voice session, the tool gets called (the agent says "One moment while I look that up") but the agent never speaks the result, and tracing shows no tool call or tool output at... Use this skill when you run into "Realtime agent goes silent after tool calls: set initialmodelsettings on .run()".

When not to use

If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.

Versions

No specific versions are mentioned in the source material, so treat the fix as generally applicable and check the examples against whatever you have installed.

Why this happens

The original report does not dig into a root cause. It documents the symptom and the fix that resolved it.

In gh:openai/openai-agents-python#1156 (OpenAI Agents SDK), the issue reported was: Issue openai/openai-agents-python#1156 (closed, 37 comments): In a RealtimeAgent voice session, the tool gets called (the agent says "One moment while I look that up") but the agent never speaks the result, and tracing shows no tool call or tool output at all, as if the tool result never reached the model. The reporter's model settings were passed to RealtimeRunner and everything looked right. The verified resolution from the thread: Reporter-verified: settings passed to RealtimeRunner are not forwarded when you call .run(). You have to set initialmodelsettings= on the .run() call itself too. Once the reporter duplicated the model settings onto .run(), the tool results flowed back into the model and tracing showed the tool calls. If your realtime agent calls tools but goes silent afterwards and the trace is empty, check that your model settings are set in both places, not just on the runner. Source: mount the host path https at container path //github.com/openai/openai-agents-python/issues/1156