v0.2.21: Fix daemon crash - reduce control loop to 2Hz, pause during audio

v0.2.20: Revert audio/satellite/voice_assistant/models/main to v0.2.9 working state

v0.2.19: Force localhost connection mode to prevent WebRTC errors

v0.2.18: Fix audio playback - restore wakeup sound, use push_audio_sample

v0.2.17: Remove head_joints/passive_joints, move error_message to diagnostic

v0.2.16: Fix TTS playback - pause recording during playback

v0.2.15: Use play_sound() instead of push_audio_sample() for TTS

v0.2.14: Pause audio recording during TTS playback to prevent GStreamer conflicts

v0.2.13: Don't manually start/stop media - let SDK/daemon manage it

v0.2.12: Disable breathing animation to prevent serial port overflow

fix: Remove unnecessary import and sleep

fix: Remove redundant import time

v0.2.11: Disable wakeup sound to prevent daemon crash + add debug logging

v0.2.10: Add debug logging for motion init + fix audio fallback samplerate

fix: Use gstreamer media_backend + add debug logging for motion init

debug: Add logging for motion and audio initialization

fix: Use push_audio_sample instead of play_sound to avoid GStreamer conflicts

v0.2.9: Remove DOA/speech detection - replaced by face tracking

fix: Fix pyproject.toml TOML syntax error

docs: Update PROJECT_PLAN.md - add face tracking, update Phase 14/17

v0.2.8: Replace DOA with YOLO face tracking

v0.2.7: Add DOA caching to prevent ReSpeaker query overload

Fix: Move _ReSpeakerContext class to correct position (before ReachyController)

Fix: Move _is_streaming_audio init to early in __init__ before entity setup

Fix: Check satellite initialization before accessing _is_streaming_audio

v0.2.6: Add thread-safe ReSpeaker USB access to prevent daemon deadlock

update readme

更新tag

Update installation steps: add 'Show community apps' checkbox step

v0.2.4: Fix microphone volume control - Use daemon HTTP API (/api/volume/microphone/set) instead of ReSpeaker DSP parameters - This matches the official implementation using amixer commands

v0.2.3: Fix daemon crash caused by conflicting pose commands - Disable pose setter methods in ReachyController (head_x/y/z, head_roll/pitch/yaw, body_yaw, antenna_left/right, look_at) - These setters conflict with MovementManager's 10Hz control loop - Per PROJECT_PLAN.md principle 7: voice conversation is highest priority - Pose control is now exclusively handled by MovementManager - Update version in __init__.py

v0.2.2: Fix second conversation motion failure - Reduce control loop from 20Hz to 10Hz - Improve connection recovery (faster reconnect, 2s interval) - Add consecutive error tracking before marking connection lost - More robust error handling for transient connection issues

v0.2.1: Fix daemon crash issue and optimize code

"revert_to_version_0.2.0"

docs: update PROJECT_PLAN and optimize motion feedback

refactor: optimize conversation motion feedback system

refactor: integrate emotion playback into MovementManager queue system

fix: make emotion playback non-blocking to prevent voice control issues

feat: add automatic facial expressions during conversation

release version 0.2.0

"fix_event_loop_blocking"

"refactor_sdk_compliance"

"fix_shutdown_handling"

"fix_connection_error_handling"

update framework

优化代码

removed wrong workflow

fix the motion

优化程序崩溃

\"chore(release): version 0.1.5\"