One voice per SI on every channel: the voices view names the voice heard on the cascaded pipeline and on the live voice model (TPSAMEVOICE)

Shipped

October 28, 2026 at 11:42 PM UTC

Author

Kamo

Commit

72500a0

speech API voices it accepts (probed) and its own; their genders are curated. The voices view gains 'effective' (speech + live voice, same or closest) and marks each option a live model also speaks ('everywhere'). kamoai-voice reads it per call. DML for crm's registry rows and its three SIs' one voice in docs/sql.