Solver

Solver 0.7.2

Chat moderation stops trusting a single AI call, staff commands get much faster to type, and the next milestone's toolkit ships early as experimental. Multi-signal corroboration for sanctions, 4 configurable levels of how much the AI gets involved, standalone commands for staff/sanctions/moderation, and (experimental) vanish/freeze/staff-chat with granular per-command permissions.

Highlights

• Experimental: Staff Mode Toolkit — /solver vanish, /solver freeze <player>, and /solver staffchat, all under /solver and also available as standalone commands (see below). Early, in-progress piece of the next big milestone, shipped now instead of waiting on it — expect rough edges.
• Experimental: Granular permissions, LuckPerms-friendly — every /solver subcommand now has its own permission node, grouped under solver.diagnostics.*, solver.sanctions.*, and solver.staffmode.*. solver.admin still grants everything, fully backward-compatible with existing OP-based setups.
• Staff/moderation commands no longer need the /solver prefix — /vanish, /mute, /ban, /moderation, and more now work directly, with /solver <command> still working exactly as before. Toggle a whole category off in config.yml if another plugin already claims one of those names.
• Chat moderation no longer sanctions on the AI's word alone — a real warn/mute/kick only fires automatically when the AI verdict is both very confident and severe, or when an independent local signal backs it up. Otherwise it stays a staff-only alert for manual review.
• 4 configurable levels of how aggressively chat moderation calls the AI — from never (local filter only, no AI provider needed) to real-time per-message.

More details: docs.jeamcube.com/en/solver/changelog