Feature: ability to post-process responses from LLM in voice assistant

[davidzhao]

Sometimes it's desirable to control or change the output from LLMs before TTS.

[seanmuirhead]

Similar: https://github.com/livekit/agents/issues/451

[parshvadaftari]

@davidzhao I'm still working on the same thing (discussed in slack), but the changes I've made in my code are specific to my requirements and not generalised. I'm not sure if that thing can help you. I'm eagerly waiting for the solution of this.

[ericyue]

@theomonnom why this marked as completed?
can we use this feature now? is there any docs about this? any information will be appreciated:)

[davidzhao]

you'd want to use the before_tts_cb callback with VoiceAssistant. you can reference this example here