@millions-hair-24956 I have class CustomVoiceAssistant(agents.voice_assistant.VoiceAssistant): I need to redeclare on llm event in CustomVoiceAssistant

@self._llm.on("metrics_collected")
def _on_llm_metrics(llm_metrics: metrics.LLMMetrics) -> None:
    speech_data = SpeechDataContextVar.get(None)
    if speech_data is None:
        sequence_id = None
    else:
        sequence_id = speech_data.sequence_id
    self.emit(
        "metrics_collected",
        metrics.PipelineLLMMetrics(
            **llm_metrics.__dict__,
            sequence_id=sequence_id,
        ),
    )

To redeclare the

on

event for LLM metrics in your

CustomVoiceAssistant

class, you can override the existing event handler by using the

@self._llm.on("metrics_collected")

decorator. This allows you to customize the behavior when LLM metrics are collected. Here's how you can implement it:

class CustomVoiceAssistant(agents.voice_assistant.VoiceAssistant):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self._llm.on("metrics_collected", self._on_llm_metrics)

    def _on_llm_metrics(self, llm_metrics: metrics.LLMMetrics) -> None:
        speech_data = SpeechDataContextVar.get(None)
        sequence_id = speech_data.sequence_id if speech_data else None
        self.emit(
            "metrics_collected",
            metrics.PipelineLLMMetrics(
                **llm_metrics.__dict__,
                sequence_id=sequence_id,
            ),
        )

This code snippet sets up a custom event handler for the

metrics_collected

event, which processes LLM metrics and emits them with additional context. Would you like more details on this implementation? Sources: - LiveKit Voice Assistant Documentation 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

@millions-hair-24956 Good, but it also keeps executing old method _on_llm_metrics from the parent class

I'm not able to find any data sources to answer your question. You might want to provide more details about how the parent class is structured or how the event system is set up. If this doesn't help, please reach out to the Livekit support team for further assistance. 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

@millions-hair-24956 I have class CustomVoiceAssistant(agents.voice_assistant.VoiceAssistant): I did

self._llm.on("metrics_collected", self._on_llm_metrics)

def _on_llm_metrics(self, llm_metrics: metrics.LLMMetrics) -> None:
    speech_data = SpeechDataContextVar.get(None)
    sequence_id = speech_data.sequence_id if speech_data else None
    self.emit(
        "metrics_collected",
        metrics.PipelineLLMMetrics(
            **llm_metrics.__dict__,
            sequence_id=sequence_id,
        ),
    )

But this handler from parent class is still executed. I need to exec on metrics collected only mine code

@self._llm.on("metrics_collected")
def _on_llm_metrics(llm_metrics: metrics.LLMMetrics) -> None:
    speech_data = SpeechDataContextVar.get(None)
    if speech_data is None:
        sequence_id = None
    else:
        sequence_id = speech_data.sequence_id
    self.emit(
        "metrics_collected",
        metrics.PipelineLLMMetrics(
            **llm_metrics.__dict__,
            sequence_id=sequence_id,
        ),
    )

I'm not able to find any data sources to answer your question. You might want to provide more details about how event handlers are registered and managed in the parent class. If this doesn't help, please reach out to the Livekit support team for further assistance. 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.