Is there a way to trigger the STT -> LLM -> TTS pipeline with an audio file? Like if I give an audio file name through an rpc call for an audio file saved in the working directory, I wanna be able to stream the audio file through the pipeline to simulate a user's speech. The use case for this is to reproducably benchmark the transcription delays and end of utterance delays with the same input audio. If I'm there's a better approach I'm missing, I'm open to ideas