🧩 FortressCore v1.0.2 — Gemini AI Integration

New: AI Security Modules (Gemini 3.5 Flash)

• AI Join Threat Analysis — profiles every player on connect, auto-kicks/bans based on threat level
• AI Chat Moderation — real-time toxicity, ad, scam, and hate speech detection with auto-actions
• AI Command Abuse Detection — detects exploit probing and permission escalation in batches
• AI Login Anomaly Detection — flags account takeover via IP region and time fingerprinting
• AI Watchlist Automation — behavioral monitoring with auto-verdict after configurable watch period
• AI Ban Appeal System — sincerity analysis with deny/reduce/pardon recommendations for owners
• AI Security Reports — scheduled threat summaries with trend analysis broadcast to all owners

New Commands

• /fc report latest / generate — AI security reports
• /fc watchlist add/remove/list — manage AI watchlist
• /fc appeal list/review/accept/deny/reduce — manage AI-assisted ban appeals

New Config (config.yml)

• Full ai: section with per-feature toggles
• Configurable Gemini prompt templates per feature
• Rate limiting (default: 15 calls/minute)
• Fallback to rule-based logic on API failure
• model: gemini-3.5-flash

Technical

• New package: me.fortress.fortresscore.ai
• GeminiAIManager — singleton HTTP client with async rate limiter and JSON parser
• AIActionExecutor — dispatches verdicts to main thread via Bukkit scheduler
• AIRateLimiter — token bucket implementation
• All AI calls fully async, zero main thread blocking
• All AI decisions logged to audit-log.yml with feature, action, latency, and prompt hash
• Gson used for all AI JSON parsing

Improvements

• /fc status now displays AI module usage stats
• Audit log extended with AI action categories
• Owner notifications for medium/high AI threat flags
• appeals.yml added for persistent appeal storage
• ai-watchlist.yml added for watchlist persistence
• security-reports/ folder for report archiving