Release v2.2.0 has the following new features:
TODO:
- Added a public pool of processors you can use; the plugin picks one automatically from the directory (by availability / load). Use PUBLIC mode in config, or CUSTOM for your own processor URL.
- Processor updates: pool mode (
pool_server), optional JSONL audit log and per-license transcript files for operators; contact me for information on setting this up, this is mainly for larger servers.- Plugin updates: connects in PUBLIC or CUSTOM mode, talks to the directory when needed.
- Deploy / install: updated processor configuration ; if you’re on the latest processor and use the egg, restart to pick up changes.
- Plugin config needs to be reset if you want to use the public processors.
- Voicechat reporting ingame
- Allow public processors to support sending audio to webhooks
- Web dashboard
- Finish up LLM processing for multiple languages
- Updated processor to support the following updates
- Added option to send voice recording that was flagged to Discord
- Overall improvements for performance
- Fixed issue with config not generating on processor
- Fixed issues with permissions
- Updated Pterodactyl egg (see processor github)
- Updated Audio queue system
- Updated VAD filtering
- Fixed max audio length configuration
- Improve overall performance of both plugin & processor
Fixes:
- Fixed Opus decoder crashes (per-player decoders)
- Fixed processing_time_ms always showing 0ms
- Fixed plugin initialization errors
Features:
- Multi-language word lists with auto-detection
- LLM profanity detection (OpenAI/Anthropic/Ollama) (still testing)
- Real-time status display
- Auto-creation of default config files
Improvements:
- Better audio buffering (complete phrases)
- Optimized LLM prompts for speed
- Fixed thread safety issues
- Updated Docker/Pterodactyl deployment files
