javis-bot d4e5e7f3f7 perf: pre-warm Whisper + chat model + TTS at bridge startup ...

The first spoken turn paid a ~10s cold start because Whisper (default
"medium") and the Ollama chat model loaded lazily on the first request.
Warm them (and ping the TTS worker) in a background thread at startup so
the server accepts requests immediately while models load, and the first
real utterance is fast.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

2026-06-14 00:04:03 +09:00

..

__init__.py

Add Discord-native hybrid front-end for Jarvis (bot + bridge)

2026-06-09 14:51:05 +09:00

melo_worker.py

feat(tts): add MeloTTS Korean voice via warm worker with offline-baked cache

2026-06-12 19:01:54 +09:00

requirements-bridge.txt

Enable GPU: LLM + Whisper on the RTX 5050, pick qwen3:8b

2026-06-09 15:49:21 +09:00

server.py

perf: pre-warm Whisper + chat model + TTS at bridge startup

2026-06-14 00:04:03 +09:00

stt_filter.py

feat(bridge): gate Whisper behind Silero VAD; harden broadcast auto-start

2026-06-13 14:53:54 +09:00

text_utils.py

fix(bridge): keep decimals, versions and URLs whole in TTS sentence split

2026-06-13 01:28:37 +09:00