- playback.zig: sentence queue + playback thread — agent reply chunks are
split on sentence boundaries and each sentence is synthesized+played as it
completes, so voice starts after the first sentence, not after the reply
- melo_server.py v2: WAV written to /dev/shm, raw s16le PCM + sample rate
streamed over stdout (WAV n rate header + bytes), file deleted immediately;
melo stdout chatter rerouted to stderr to keep the protocol clean
- melo.zig: speakRaw returns in-memory PCM; pw-play fed via stdin pipe
(no on-disk TTS files at all)
- whisper: --language auto
- fix: double-close of pw-play stdin panicked Threaded Io in debug
- fix: quit-path frees for transcript cache, bubble lists, sentence queue,
reply buffer (debug allocator leak panic)
👾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta Code <noreply@letta.com>
- melo_server.py subprocess (MeloTTS v3 EN-Newest, speed 1.3): line protocol
stdin->stdout, model loaded once, warmup before READY
- src/melo.zig: lifecycle ownership (spawn/kill with inferon), line protocol
client, reply sanitizer (newlines/markdown stripped before synthesis)
- SIGINT/SIGTERM handlers: async-signal-safe kill of whisper/melo/recorder,
default disposition restored + re-raise for truthful exit status
- processUtterance: transcribe->infer->speak chains automatically after
stop-click; mid-pipeline clicks are no-ops; back to idle after reply
- temp cleanup: recording WAV deleted after transcription, TTS WAV after
playback
- whisper: pid() export for signal handler
👾 Generated with [Letta Code](https://letta.com)
Co-Authored-By: Letta Code <noreply@letta.com>
pierreandLetta Code