Increase hey_reachy probability_cutoff to 0.9 to reduce false triggers

v0.5.11: Reset feature extractors when switching wake words to prevent false triggers

v0.5.10: Actually fix wake word switching - set id attribute on loaded models

v0.5.9: Default wake word to hey_reachy, fix wake word switching bug

v0.5.8: Fix tap detection during emotion playback - poll daemon API for move completion

fix: Disable tap detection during emotion playback to prevent false triggers

fix: Add debug logging for noise suppression, fix PP_MIN_NN value

fix: EntityRegistry has no 'state' attribute - use server.state

fix: Invert DOA yaw direction - was turning opposite way

fix: Show raw DOA angle in Home Assistant (0-180)

fix: Change DOA turn-to-sound logs to INFO level for visibility

fix: Remove speech_detected check in DOA turn-to-sound

feat: Add DOA turn-to-sound at wakeup (v0.5.7)

chore: bump version to v0.5.6

fix: better pipeline state tracking to prevent duplicate audio

chore: bump version to v0.5.5

fix: prevent concurrent pipelines and add prompt sound for continuous conversation

chore: bump version to v0.5.4

fix: wait for RUN_END before starting new conversation

chore: bump version to v0.5.3

fix: improve continuous conversation with conversation_id tracking

chore: bump version to v0.5.2

fix: enable HA control of robot pose and fix continuous conversation

fix: update emotion mapping to use available robot emotions

docs: rename project to 'Reachy Mini for Home Assistant'

fix: send stop request to HA when exiting tap conversation mode

fix: allow tap to exit conversation without refractory delay

docs: fix STT/TTS engine description - user configured in HA

docs: update Technical Architecture diagram with full system overview

docs: translate PROJECT_PLAN.md to English (complete)

docs: translate PROJECT_PLAN.md to English

docs: update changelog for v0.5.1 bug fixes

修复: Sendspin连接到music_player而非tts_player

持久化tap_sensitivity设置

修复: tap_sensitivity默认值改为0.5g, 修复camera_server属性名

修复: 语音助手唤醒时暂停Sendspin和音乐播放

Sendspin: 优先使用16kHz采样率(ReSpeaker原生支持)

更新文档: 同步人脸追踪资源优化功能

优化人脸追踪资源使用: 自适应帧率

人脸模型下载加重试机制(3次重试，间隔5秒)

恢复到4051ebd: 手势识别有依赖冲突，暂时移除

mediapipe依赖移到pyproject.toml，代码只安装mediapipe本身

手势识别: 首次运行自动安装mediapipe 0.10.18 (ARM64)

手势识别: 移除自动安装，无mediapipe时静默禁用

手势识别: ARM64自动安装mediapipe 0.10.18 (--no-deps避免numpy冲突)

手势识别改用纯OpenCV (ARM兼容，无MediaPipe依赖)

更新文档: 手势识别增加到11种

v0.6.0: MediaPipe手势识别(11种手势)

手势识别增加到11种: thumbs_up/down, open_palm, fist, peace, pointing_up, ok, rock, call, three, four

手势识别改用MediaPipe (0.10.31+支持numpy2)