senator-sperma/cogs, branch main

OPEN THE RUST GATES

2026-07-08T00:13:35Z

Signed-off-by: salmonik

Merge pull request 'Voice related fixes, alongside simplifying assistant' (#5) from dev into main

2026-05-21T15:10:23Z

Reviewed-on: https://codeberg.org/salmonik/senator-sperma/pulls/5

Fix voice reconnect race: reuse VoiceClient instead of destroying it

2026-05-21T14:45:31Z

The stale-connection path was calling _remove_voice_client() then creating a brand-new VoiceClient via channel.connect(). This created a race: VOICE_SERVER_UPDATE (triggered by Discord when the user joins voice) arrives between the destroy and the fresh-connect, gets consumed by the old client, and the new client never receives it — causing a 10s timeout waiting for got_both_voice_updates. Fix: after disconnect(), call vc.connect() on the same VoiceClient instance. The disconnect resets the state machine to 'disconnected' but the same VoiceConnectionState processes any events that arrive between disconnect and reconnect, so the state machine correctly reaches got_both_voice_updates.

make the assistant simpler

2026-05-21T14:38:01Z

Signed-off-by: salmonik

Fix zombie voice client after gateway restart

2026-05-20T15:19:11Z

- play_tts: run is_voice_client_alive() check regardless of is_playing() state. After a gateway restart, is_playing() can return True on a dead client, which previously bypassed the stale-detection guard (was: 'if is_connected and not is_playing')

Fix stale-connection reconnect hang and TTS temp-file collision

2026-05-20T15:08:40Z

- _ensure_voice_client: call _remove_voice_client() when cleaning up stale or disconnected clients so the fresh connect doesn't hang for 30s waiting for Discord's stale state to clear - Reduce connect timeout from 30s to 10s for faster failure recovery - vclogs.py: use a unique temp filename per TTS call (include timestamp) instead of relying on PID alone, preventing concurrent TTS calls from clobbering each other's files

Redirect play_commands imports to cogs.voice, remove audio_playback shim

2026-05-20T14:40:28Z

- Update play_commands.py to import get_voice_manager from cogs.voice instead of going through the audio_playback.py re-export shim - Delete the audio_playback.py shim file entirely - Update test_music_audio.py imports to use cogs.voice.VoiceManager - Delete PLAN_VOICE_REFACTOR.md (plan is complete)

Clean up consumers: replace polling with VoiceManager.play_source()

2026-05-20T13:40:18Z

- cron.py: replace while-is_playing polling with VoiceManager.play_source() - vclogs.py: replace while-is_playing polling with VoiceManager.play_source() - stop.py: use VoiceManager.disconnect() instead of raw voice_client - join.py: use VoiceManager.connect() instead of raw channel.connect() - voice.py: add back join_by_channel_id() convenience method - Tests: update for new VoiceManager method names and signatures

Fix TTS and AI voice playback bugs

2026-05-20T13:37:14Z

- tts.py _handle_voice_output: add missing voice_client.play() call via VoiceManager.play_source() — was creating the audio source and then doing nothing - ai.py: use VoiceManager.play_source() instead of passing an FFmpegPCMAudio object to play() which expects a string path

Fold AudioPlaybackManager into VoiceManager + Future-based play

2026-05-20T13:36:17Z

- Move all static helpers (validate_voice_channel, prepare_source, prepare_audio_source, etc.) from AudioPlaybackManager into VoiceManager - Add _play_await() using asyncio.Future + after callback, replacing all polling patterns and the 5-second sleep hack - Add play() high-level method and play_source() for pre-made sources - Unify connection logic under single _ensure_voice_client() - Remove dead Voice cog class and its setup() - Gut audio_playback.py to a re-export shim - Remove cogs.voice from main.py extensions