Voice

Voice settings for chat, and a push-to-talk session against one agent. Transcripts are DLP-redacted before the model.