added more logging around $YESNO so we can get more consistent yes and no replies processed.... they are spotty currently

This commit is contained in:
Jacob Dubin
2026-05-03 22:42:41 -05:00
parent 2ec4902189
commit 573911de0f
8 changed files with 185 additions and 34 deletions

View File

@@ -207,6 +207,7 @@ Expected:
- no `ffmpeg` failure should become the dominant failure mode for non-Opus buffered audio
- short replies such as `yes`, `no`, `cancel`, and short alarm times should either map correctly or be classified as STT misses with evidence
- when chasing a flaky `$YESNO` reply, look for the new turn telemetry categories `binary_audio_received`, `binary_audio_ignored`, `yes_no_turn_received`, `yes_no_turn_resolved`, and `yes_no_no_input`; the useful question is whether the short reply still had `AwaitingTurnCompletion = true`, active listen rules, and buffered audio when it hit the finalizer
### Stop And Volume