Describe the bug

According to the documentation https://docs.livekit.io/agents/models/realtime/#separate-tts
it should be possible to use a realtime model with a different TTS instance.

But it seems the modalities paramter is not implemented yet, there are some parts int he code that mention it already, but the implementation itself is missing.

Unfortuanteliy I tried to "hardcode" it and modified the library locally.. even though I now explicity set "text" to modalitiy, it's still not working. ... I guess something else is missing here?

Relevant log output

No response

Describe your environment

"name": "@livekit/agents-plugin-openai",
"version": "1.0.11",

Minimal reproducible example

No response

Additional information

No response