 wenhongquan
|
ffb8cc2d1a
feat(session): 实现房间恢复功能，增加会话服务以管理最后使用的房间
|
há 3 semanas atrás |
 wenhongquan
|
e1a9225a6e
feat(memory): two-tier memory - persistent voiceprint + session context
|
há 3 semanas atrás |
|
wenhongquan
|
2207bfa3fc
fix(memory): use JSON file for conversation memory (Zvec can't do scalar-only query)
|
há 3 semanas atrás |
|
wenhongquan
|
f7486333df
feat(memory): speaker-aware conversation memory via Zvec
|
há 3 semanas atrás |
|
wenhongquan
|
d5d1c71fb5
feat(voiceprint): auto-register on first speech, hide speaker name from LLM output
|
há 3 semanas atrás |
|
wenhongquan
|
4abc018a4f
feat(voiceprint): Zvec-backed voiceprint registration + identification
|
há 3 semanas atrás |
|
wenhongquan
|
d3e2c18286
fix(tts): only interrupt TTS on real ASR, filler detection won't stop playback
|
há 3 semanas atrás |
|
wenhongquan
|
8f4d80c99a
debug: add PLAY start/end logs for _player
|
há 3 semanas atrás |
|
wenhongquan
|
91f8020278
debug: add LLM start/done logs to trace TTS trigger
|
há 3 semanas atrás |
|
wenhongquan
|
6e57864dd8
feat(worker): auto-exit when room empty, cleanup task
|
há 3 semanas atrás |
|
wenhongquan
|
0f736e4761
perf(worker): preload VAD at startup, skip Qwen3 when using Mimo ASR
|
há 3 semanas atrás |
|
wenhongquan
|
c92b1edf2c
cleanup: merge worker files, remove simplified version
|
há 3 semanas atrás |
|
wenhongquan
|
1059df76bc
fix(worker): use non-overlapping transcription window to eliminate duplicate outputs
|
há 3 semanas atrás |
|
wenhongquan
|
a221c1ae02
fix(livekit): add participant_connected and track_published handlers to fix track_subscribed not firing
|
há 3 semanas atrás |
