ВЕРСИИ (12)

v3.9 — Changelog Critical Fixes - RL learning now actually works. A key mismatch between how the AI tracked responses and how it scored them meant the reinforcement learning system was effectively doing nothing since v3.8. Fixed — Pokemon now genuinely learn which responses work best over time. - Response variety massively increased. A probability bug was squaring all content gate chances (a 30% chance was really 9%, a 5% chance was 0.25%). Fixed across 48 gates — Pokemon now use their full range of personality expressions, comments, and reactions. - Evolution responses no longer get cut off. Evolution is now treated as a critical event that bypasses the response budget limiter. AI Quality - Smarter learning integration. The system that scores how much learning data is available was penalizing Pokemon for having more knowledge sources. Fixed — Pokemon with richer learning history now give better responses. - Knowledge graduation no longer wipes Pokemon memory. When shared knowledge was promoted to the global pool, it was being deleted from individual Pokemon. Fixed — your Pokemon keep everything they've learned. - No more duplicate reward tracking. The RL engine was double-counting rewards, making the learning signal noisy and unreliable. Fixed. - Exploration/exploitation balance stabilized. Negative feedback was pushing Pokemon toward random responses twice as fast as positive feedback pulled them back. Now balanced — Pokemon recover smoothly after bad interactions. - Italian sarcasm detection fixed. Expressions like "siiii!" and "noooo!" (genuine Italian enthusiasm) were being flagged as sarcasm. Fixed — only actual sarcastic elongations trigger detection now. - External AI learning data no longer cross-contaminates. Usage stats from one lookup category were bleeding into others, skewing diversity penalties. Fixed. Stability - Save system is more resilient. If an async save fails, the mod now correctly retries next cycle instead of silently skipping it. - Relationship cleanup uses exact matching. Prevents edge cases where resetting one Pokemon could affect another

● v3.8 Changelog 1. Commands.kt Full Bilingual (~112 strings) - All 22 command handler functions now support English via Lang.t() - Labels: Species, Level, Friendship, Weather, Ability, Types, Status - Relationship types: BEST FRIENDS, FRIENDS, ACQUAINTANCES, NEUTRAL, DISLIKE, RIVALS - Time of day: Morning, Day, Sunset, Night, Dawn - Weather: Storm, Rain, Clear - All help text, error messages, admin commands, friendship bonuses, battle sync, world context 2. RL Action Key Fix (CRITICAL BUG) - trackAction() was storing literal string "response" as action key since v2.4 - All policies had only 1 action entry → RL scoring in SRG used part.take(40).lowercase() as lookup key → never matched "response" → getActionSuccessProbability() always returned default 0.5f - The entire RL scoring boost was non-functional since inception - Fix: now stores finalResponse.take(40).lowercase() as action key, matching what scoring looks up 3. RL-Driven Response Generation (NEW) - New generation path using chooseBestAction() to produce response parts from learned policies - Guards: 30+ episodes (no cold start), action 5-150 chars, not old "response" key, success probability > 60%, budget-gated at 15% - Tracked as "rl_response" feature source for feedback loop 4. Context Quality Threshold Lowered - SRG threshold: 0.5f → 0.35f (0.5 was blocking learning context for new/young Pokemon) - LCI: added enhancementScale — at quality 0.35 → 30% enhancement strength, at 0.5+ → 100% - New Pokemon now benefit from learning data much earlier 5. LanguageGrowth Catchphrase Getter - Added public getCatchphrase(pokemonId): String? method - Catchphrases were already used internally in enrichResponse() at 5% chance, but now accessible externally 6. ExternalAI Diversity Penalty Fix - Old: sqrt(excess) * 0.03 subtracted — response used 100 times still had ~70% quality - New: ln(1 + excess) * 0.06 multiplicative — 10 excess: ×0.86, 50: ×0.76, 100: ×0.72, 500: ×0.63 - Floor at 5% prevents total data loss 7. TPS Optimizations - Cache TTL: CACHE_TTL_TICKS 10 → 40 (0.5s → 2s) — event-based invalidation handles fast updates, TTL is just a fallback - GlobalKN flatMap: flatMap { listOf() + list }.toSet() → mutableSetOf() + forEach { addAll } — eliminates intermediate List allocation - GlobalKN eviction: sorts lightweight Pair<String, Float> instead of full Map.Entry objects

CRZbrain v3.7 — Bugfix + Thread Safety + Performance + AI Quality Thread Safety & Memory Leaks (Batch 1) TickOptimizationSystem.kt - Fix: Replaced global lastWorldEventTick / lastRelationshipTick counters with per-player ConcurrentHashMap<UUID, Long> — previously only the first player in the loop would trigger world events each cycle, causing uneven event distribution across players - Fix: Added cleanupPokemon(pokemonId) method to remove stale entries from pokemonDistanceTier map (was never cleaned, leaked ~1KB/Pokemon/hour) - Fix: Added cleanupPlayer(playerUuid) method to remove per-player tick counters and cache on disconnect AdaptiveLearning.kt - Fix: Wrapped LRU eviction of trainerPatterns in synchronized block with double-check pattern — concurrent modification could corrupt the map during eviction - Fix: Added dead zone handling for success rates between 0.3–0.6 — previously no preference was recorded for ambiguous responses, now both liked/disliked decay toward zero - Fix: Changed context matching from exact equals(ignoreCase=true) to fuzzy word-overlap matching (≥50% overlap) — strategies learned for "battle_wild" now also apply to "battle_wild_grass" - Fix: Improved duplicate strategy detection key for more reliable dedup IntelligenceEvolution.kt - Fix: Wrapped recentResults ArrayDeque access in synchronized(stats) block — ArrayDeque is not thread-safe, concurrent addLast/removeFirst could throw ConcurrentModificationException - Fix: Limited word association pairs to max 8 random samples for large word lists (was O(n×5) linkConcepts calls for every message) — also capped context associations to 3×3 entries LearningContextIntegration.kt - Fix: Added @Volatile annotation to contextWeightsDirty flag — without it, the JVM could cache the value in a CPU register and the save thread would never see updates from chat threads DeepUnderstanding.kt - Fix: Replaced single-entry LRU eviction with bulk 20% removal when playerCommunicationPatterns exceeds 200 entries — was removing 1 entry per call, causing O(n) overhead every call after cap - Fix: Capped context string in learnContextualMeanings to 100 characters — unbounded joinToString was producing very long map keys, wasting memory - New: Added saveData() / loadData() persistence for playerCommunicationPatterns and learnedMeanings to config/CRZbrain/learning/deep_understanding.json — previously all learned communication styles and word meanings were lost on server restart CRZbrainMod.kt - Integration: Added DeepUnderstanding.loadData() call in SERVER_STARTED handler - Integration: Added DeepUnderstanding.saveData() call in SERVER_STOPPING handler - Fix: Added TickOptimizationSystem.cleanupPlayer() call on player disconnect alongside existing invalidateCache() PokemonLifecycleManager.kt - Fix: Added TickOptimizationSystem.cleanupPokemon(pokemonId) call in cleanupRamOnlyData() to remove stale distance tier entries when Pokemon are released or removed --- Performance (Batch 2) ReinforcementLearning.kt - Fix: Replaced O(n log n) sortByDescending in replay buffer eviction with O(n) minimum priority scan — buffer was being fully sorted on every single add just to remove one lowest-priority entry ExternalAILearning.kt - Fix: Removed unnecessary .copy() calls for event/emotion/topic maps in learnFromResponse() — was duplicating every LearnedResponse 5 times (one per index), now shares same object reference for nature/event/emotion/topic (species keeps copy due to different eviction lifecycle). ~60% memory reduction on learned responses - Fix: Changed overuse penalty from linear (count - threshold) * 0.05f to sublinear sqrt(count - threshold) * 0.03f — linear penalty was killing responses after ~20 uses, sqrt curve is much gentler (4 over → -0.06, 16 over → -0.12, 100 over → -0.3) - Fix: Capped feedbackScore at [-10, 10] and feedbackCount at 50 — previously accumulated without bound, causing responses with many feedbacks to dominate all quality calculations AdvancedPatternMining.kt - Fix: Replaced message.hashCode().toString() document ID with AtomicLong counter + System.nanoTime() — 32-bit hashCode collisions caused different messages to overwrite each other's topic distributions - Fix: Capped initial topic keywords to 50 on creation (was unbounded from long messages) GlobalKnowledgeNetwork.kt - Fix: Replaced hash-based strategy ID "${context.hashCode()}_${strategy.hashCode()}" with string concatenation "${context}__${strategy}" — hash collisions caused different strategies to merge - Fix: Skip trainer diversity penalty when totalTrainers < 3 — with 1-2 players, diversity ratio was artificially low (0.3-0.5), unfairly penalizing consensus scores - Fix: When global knowledge map is full, evict entry with lowest consensus score instead of silently rejecting new knowledge — previously new valid knowledge was dropped once cap was reached --- AI Quality & Bug Fixes (Batch 3) SmartResponseGenerator.kt - Fix: Removed dead playerMessage != null null check at line 671 — playerMessage is a non-null String parameter, check was always true (compiler warning) - Fix: Removed dead playerMessage != null null check at line 2266 — same issue, second occurrence. Both pre-existing compiler warnings are now resolved (0 warnings) - Fix: Consolidated double playerMessage.split() into single cached split, reused for both learnConcept and linkConcepts calls - Fix: Replaced java.time.LocalDateTime.now().hour fallback with constant 12 (noon) — game time ≠ system time, using real clock for unrecognized timeOfDay strings was producing wrong learning context - Fix: Added division-by-zero guard for maxHp in HP percentage calculation — maxHp == 0 (eggs, data glitches) produced Infinity which toInt() converted to Int.MAX_VALUE - Fix: Raised learning context quality threshold from > 0.3f to > 0.5f — low-confidence context was degrading response quality by applying unreliable enhancements MoveAnimationSystem.kt - Fix: Added distance check startPos.squaredDistanceTo(endPos) < 0.01 before normalize() in playSignatureAnimation — when attacker and target are at the same position, normalizing a zero vector produces NaN components, causing particle positioning to fail silently --- Files Modified (13) 1. sensors/TickOptimizationSystem.kt 2. ai/embedded/AdaptiveLearning.kt 3. ai/embedded/IntelligenceEvolution.kt 4. ai/learning/LearningContextIntegration.kt 5. ai/embedded/DeepUnderstanding.kt 6. CRZbrainMod.kt 7. PokemonLifecycleManager.kt 8. ai/learning/ReinforcementLearning.kt 9. ai/learning/ExternalAILearning.kt 10. ai/learning/AdvancedPatternMining.kt 11. ai/learning/GlobalKnowledgeNetwork.kt 12. ai/embedded/SmartResponseGenerator.kt 13. combat/MoveAnimationSystem.kt New Persistence File - config/CRZbrain/learning/deep_understanding.json — stores player communication patterns and per-Pokemon learned word meanings

v3.6: Defense System Disabled + Per-Pokemon Tick Optimization Part A — PokemonDefenseSystem Disabled 1. Removed PokemonDefenseSystem.register() from CRZbrainMod — no more tick handler (entity search every 1s for ALL Pokemon) and no damage listener 2. Removed all PDS calls from PokemonCommandSystem — executeProtect(), executeActiveDefense(), executeAggressiveMode(), executeGuardMode(), executeStandDown() no longer call PokemonDefenseSystem 3. Simplified updateDefendingState() — removed findHostilesNearPlayer() entity search, now just follows player (teleport if distance > 5.0) 4. Simplified updateProtectingState() — removed findHostilesNearPlayer() entity search, now just stays near player (teleport if distance > 3.0) 5. Disabled defense commands in Commands.kt — setDefenseMode(), showDefenseStatus(), showAllDefenseStatus(), activatePvPMode() all return "Sistema difesa disabilitato (v3.6)" 6. Removed PDS.resetPokemon() from PokemonLifecycleManager 7. Removed unused imports — PokemonDefenseSystem import removed from CRZbrainMod, Commands, PokemonCommandSystem, PokemonLifecycleManager Part B — TickOptimizationSystem (New Unified Tick Dispatcher) 8. Created sensors/TickOptimizationSystem.kt — single ServerTickEvents.END_SERVER_TICK handler replaces 3 separate handlers 9. Strategy 1 — Temporal Staggering: Pokemon split into 2 groups via uuid.hashCode() % 2, updated on alternating ticks (halves peak load per tick) 10. Strategy 2 — Distance-Based LOD: Pokemon >16 blocks from trainer (squaredDistanceTo > 256) skip every other update cycle (recalculated every 20 ticks) 11. Strategy 3 — Frequency Reduction: Combat states stay at 10-tick interval, non-combat states (FOLLOWING, EXPLORING, DEFENDING, PROTECTING) doubled to 20 ticks; WorldEventSystem interval doubled from 600 to 1200 ticks (30s → 60s) 12. Strategy 4 — Intelligent Batching: getActivePokemon() cached per player with 10-tick TTL (was called ~28 times per cycle, now 1); single player iteration loop instead of 3 separate loops 13. Removed tick handler from PokemonCommandSystem — exposed updateSinglePokemon() public method, removed lastUpdateTick and UPDATE_INTERVAL 14. Removed tick handler from WorldEventSystem — renamed checkWorldEvents() to checkWorldEventsForPlayer() and made it public, removed lastCheckTick and CHECK_INTERVAL_TICKS 15. Removed tick handler from PokemonRelationshipSystem — added tickForPlayer() public method, removed lastInteractionTick and INTERACTION_COOLDOWN_TICKS 16. Cache invalidation hooks — TickOptimizationSystem.invalidateCache() called on POKEMON_SENT_POST and DISCONNECT events; cleanup() called on SERVER_STOPPING 17. handleSpontaneousDialogue() uses cached Pokemon list instead of calling PokemonUtils.hasActivePokemon() directly 18. Removed unused imports — ServerTickEvents removed from WorldEventSystem, PokemonRelationshipSystem; HostileEntity/AnimalEntity/PlayerEntity removed from PokemonCommandSystem

v3.5: Knowledge Graduation + SRG Optimization + Auto-Pruning + File Consolidation 1. SRG Lazy Evaluation (SmartResponseGenerator.kt) - Problem: 35+ subsystems were called BEFORE the probability gate check. With average gate ~9%, 91% of subsystem calls were wasted CPU. - Fix: Inverted all probability gates so Random.nextFloat() < probability is checked BEFORE calling the subsystem. If the random check fails, the subsystem is never called. - Replaced parts.add() with parts.addPart() that tracks a partsCollected counter. 2. SRG Part Budget System (SmartResponseGenerator.kt) - Problem: 30-50 response parts were generated per response, but only 2-3 were used. - Fix: budgetGate(probability) lambda that enforces limits: - After 8 parts collected: skip any gate with probability < 15% - After 12 parts collected: skip ALL remaining gates - Priority gates (critical health, shiny, near death, evolution) bypass the budget entirely 3. Auto-Pruning of Stale Data (LearningSystem.kt + ExternalAILearning.kt) - Problem: Learning data accumulated indefinitely, files grew without bound. - Fix: pruneStaleData() runs every 5 minutes and removes: - Associations with strength < 0.1 - Behavior patterns with < 3 observations AND older than 30 days - Lessons with importance < 3 AND older than 60 days - Pokemon knowledge with confidence < 0.3 AND not reinforced in 30 days - Terminology with frequency == 1 AND older than 14 days - ExternalAI responses with effectiveQuality < 0.1 - Backward compatible: pre-v3.5 data (createdAt == 0) is never pruned by age 4. ExternalAI File Consolidation (ExternalAILearning.kt) - Problem: 11 separate JSON files for ExternalAI learning data. - Fix: Consolidated to 3 files: - learned_responses.json (byNature + byEvent + byEmotion + byTopic + bySpecies + speciesLastUsed) - learned_phrases.json (phrases + expressions + openers + closers + conversationPatterns) - stats.json (unchanged) - Backward compatible: loads old 11-file format if new format not found, migrates on next save, then deletes old files 5. Knowledge Graduation System (LearningSystem.kt) - Problem: Each Pokemon learns independently from scratch. Data accumulated per-Pokemon without cross-pollination. - Fix: graduateKnowledge() runs every 5 minutes and promotes validated knowledge: - Vocabulary appearing in 3+ Pokemon → shared vocabulary - Associations in 3+ Pokemon with avg strength > 0.5 → shared associations - Behavior patterns with successRate > 0.75 and 10+ observations in 2+ Pokemon → shared patterns - Knowledge with confidence > 0.8 and 5+ reinforcements in 2+ Pokemon → shared knowledge - Battle strategies from Pokemon with 20+ battles → shared battle meta - Graduated entries are removed from per-Pokemon data (files shrink over time) - bootstrapNewPokemon(): Pokemon with < 10 vocabulary items automatically receive graduated knowledge as a starting base - Saved to graduated_knowledge.json Files Modified ┌───────────────────────────┬────────────────────────────────────────────────────────────────────┐ │ File │ Changes │ ├───────────────────────────┼────────────────────────────────────────────────────────────────────┤ │ SmartResponseGenerator.kt │ Lazy evaluation (35+ gates inverted) + Part budget system │ ├───────────────────────────┼────────────────────────────────────────────────────────────────────┤ │ LearningSystem.kt │ Knowledge graduation + auto-pruning + createdAt fields + version 5 │ ├───────────────────────────┼────────────────────────────────────────────────────────────────────┤ │ ExternalAILearning.kt │ File consolidation (11→3) + pruning │ ├───────────────────────────┼────────────────────────────────────────────────────────────────────┤ │ CRZbrainMod.kt │ Periodic graduation/pruning trigger (every 6000 ticks) │ └───────────────────────────┴────────────────────────────────────────────────────────────────────┘ Impact - CPU: ~91% reduction in subsystem calls, max 12 parts instead of 30-50 - Disk: Files auto-shrink via graduation + pruning, 11 files → 3 for ExternalAI - AI Quality: New Pokemon inherit validated cross-Pokemon knowledge instead of starting from zero - Build: BUILD SUCCESSFUL, only 2 pre-existing warnings (playerMessage != null)

v3.4: Bug Fix + Dead Code Activation + Full Bilingual Completion Group A: Critical Bug Fixes + Dead Code Activation (4 fixes) 1. timeOfDay always returned null (SRG.kt) — context.timeOfDay contains strings like "morning"/"night", but the code called .toIntOrNull() which always returned null, defaulting to hour=12. AdaptiveLearning's temporal patterns (learning when players are active) were completely broken. Fixed with a when mapping strings to representative hours. 2. adaptResponse() never called (SRG.kt) — AdaptiveLearning had a fully implemented function that categorizes responses and modifies unsuccessful ones based on learned preferences. It was never called from anywhere. Now inserted in SRG's post-processing pipeline. 3. isPreferred()/isAvoided() never called (SRG.kt + AdaptiveLearning.kt) — These functions check what response styles a Pokemon has learned to prefer or avoid. The data was being collected but never used. Now integrated into prioritizeResponseParts scoring: avoided categories get -15, preferred get +10. Also made adaptResponse() write response-shape categories (enthusiastic_short, questioning, hesitant, verbose, brief, standard) so preferences accumulate by response shape. 4. AdaptiveLearning bilingual gaps — predictNextTrainerAction() and generatePreferenceComment() were Italian-only. Wrapped with Lang.t(). Group B: Performance + Persistence + ExternalAI (5 fixes) 5. Per-call Regex in LearningSystem — Emoji regex was created inside updatePlayerPersonalityProfile() on every message. Hoisted to object-level private val EMOJI_REGEX. 6. GlobalKnowledgeNetwork non-atomic writes — All 6 data files used direct File.writeText(). A crash during write = data corruption. Added atomicWrite() helper using temp file + rename pattern. 7. ExternalAILearning bilingual gaps — Added 12 English interjections to extractExpressions(), 5 English question starters to extractConversationPatterns(), and an English self-reference check in evaluateQuality(). 8. AIHandler removeAt(0) — Replaced O(n) while-loop with bulk subList(0, excess).clear(). 9. PokemonDefenseSystem removeAt(0) — Same fix for target memory trimming. Group C: Bilingual IntelligenceEvolution + LCI (~143 strings) 10. IntelligenceEvolution — 31 strings across inductive/abductive/predictive inference, philosophical/analytical/pattern/learning comments, knowledge enrichment, and association recall. 11. LearningContextIntegration — 112 strings across behavior prediction, tone adaptation, player references, battle insights, strategy phrases, and 8 emotion-response generators (comfort, joy, calming, affection, reassurance, supportive, playful, serious). Group D: Bilingual EmotionalMomentum + LanguageGrowth (~252 strings) 12. EmotionalMomentum — 160 strings: all 11 moods (EXCITED, SAD, LOVING, SCARED, PROUD, TIRED, CURIOUS, AGGRESSIVE, HAPPY, LONELY, CONFUSED) with nature variants, plus modifyResponseByState() (happiness additions, yawn prefix, energy additions, curiosity additions, confidence word replacements, sigh prefix, stress truncation, affection suffix). 13. LanguageGrowth — 92 strings: emotive expressions across 5 complexity levels × 4 emotions, personal expressions by 4 nature categories × 5 levels, catchphrases by 4 categories, vocabulary upgrades using Lang.tMap() for IT→fancy-IT / EN→fancy-EN replacements, and acquired word usage templates. Group E: Bilingual LearningSystem (~200 strings) 14. LearningSystem — 200 strings across 19 functions: battle insights, learned phrases, popular topic comments, player expression usage, player relation comments, event mentions, emotional atmosphere comments, personalized greetings (5 player types × natures), player style adaptation, activity time comments, topic-based responses, player vocabulary usage, previous interaction references, learned response generation (vocabulary/associations/lessons), battle-related comments, fully personalized responses, growth comments, battle predictions, and known player greetings.

LearningSystem.kt has been modified

● v3.2 Fixes - AI Quality + Performance + Persistence (~55 fixes across 17 files) Critical: RL + Persistence (7 fixes) 1. RL: Fixed REINFORCE score function - Was advantage * (1-probability), now (advantage / probability).coerceIn(-3, 3) - same fix in experience replay 2. RL: EMA baseline - Was cumulative average (stale after thousands of episodes), now 0.99*avg + 0.01*reward 3. RL: O(n) chooseBestAction - Was O(n²) recomputing totalWeight per action inside maxByOrNull 4. RL: Thread-safe stats - totalUpdates/totalRewardsPositive/totalRewardsNegative changed to AtomicInteger; reset() now also clears replayBuffers 5. AdaptiveLearning: Full persistence - Was completely lost on restart! Now saves/loads 6 JSON files in config/CRZbrain/adaptive/ 6. AdaptiveLearning: Fixed importStrategy - Floor was 0.225 (below failure threshold), now maxOf(rate * 0.3, 0.4). Also fixed equals vs contains for context matching, capped trainerPatterns at 200 7. CRZbrainMod: Shutdown saves 4 missing systems - Added ExternalAILearning, EmotionalMomentum, LearningContextIntegration, AdaptiveLearning to shutdown. Added LearningSystem.forceSave() to periodic async saves High: Bug Fixes (14 fixes) 8. SRG: Atomic read-and-remove for feedback attribution - Was race condition where another thread could read null between get() and remove() 9. SRG: English keywords added to all scoring/filtering - Was Italian-only, English-speaking players got random responses instead of contextual ones 10. SRG: Fixed part.first() crash - Changed to firstOrNull()?.isUpperCase() == true. Added probability gates to shiny/seasonal/first-Pokemon responses 11. SRG: Cross-Pokemon RL transfer - Filter was using contains (impossible match on context strings), fixed to startsWith 12. ExternalAILearning: English keywords - Added EN keywords to detectTopic (10 categories) and detectEmotion (6 categories) 13. ExternalAILearning: Atomic writes + multiple fixes - All 11 files now use temp+rename atomic writes. Fixed diversityScore Gson deserialization. Fixed extractOpener while-loop. Added read-time quality decay 14. DeepUnderstanding: Fixed 4 broken Regex - Escaped :) :( >:) special characters. Elongation pattern now requires 2+ repeated chars 15. DeepUnderstanding: Fixed detectEmotion double-call - Was computing emotion twice per message. Added caps: 500 words/Pokemon in learnedMeanings, 200 players in communicationPatterns 16. IntelligenceEvolution: Fixed totalKnowledge - Wasn't decremented after knowledge eviction. Fixed learningSpeed cap at 3.0 using maxOf instead of overwrite 17. IntelligenceEvolution: Word-boundary Regex - "male" and "no" were matching inside other Italian words. Skip English context keys in enrichment. Minimum 2-word overlap for deductive inference 18. LCI: Thread-safe inner maps - Changed mutableMapOf() to ConcurrentHashMap() inside getOrPut calls. Fixed quality score normalization (divide by totalWeightUsed) 19. LCI: Feedback detection - Now requires keyword AND sentiment agreement. Was false-positive on common Italian words like "sono", "pokemon" 20. APM: Fixed substring matching - Sentiment markers used substring contains (false positives). Now split into single-word Set checks + explicit multi-word contains 21. EmbeddedAIProvider: Moods actually work - EXCITED was a no-op, ANGRY destroyed ellipses, TIRED never triggered. All 4 maps changed to ConcurrentHashMap Medium: Performance (14 fixes) 22. APM: Eliminated ~200 allocations per call - All per-call maps, lists, sets, regex moved to object-level vals 23. APM: Sentiment cache String key - Was using hashCode (collisions). Added caps: 100 topics, 200 actions. synchronizedList for inner lists. Atomic file writes 24. LearningSystem: Removed sync I/O from markDirty - Was calling checkAutoSave() which could block main thread. Pre-compiled 17 regex 25. SmartConversation: Pre-compiled 20+ regex - normalizeMessage was creating new Regex objects per call. Hoisted SPECIES_QUIRKS map to object level 26. MoveAnimationSystem: Cached getNearbyPlayers - Was doing 16 entity searches per attack, now does 2 (query once, pass list to all sub-functions) 27. PokemonDefenseSystem: Cached entity search - One search per player shared across all Pokemon. Replaced sortedWith().first() with minWithOrNull() 28. MemorySystem: Bulk eviction - removeAt(0) in while loop replaced with subList(0, excess).clear() (single array shift) 29. LanguageGrowth: Word-boundary replacement - upgradeVocabulary now uses Regex word boundaries to avoid replacing inside other words 30. EmotionalMomentum: Fixed stale lastUpdate - Added timestamp update in applyEmotionalTransfer and applyBuildup (was never refreshed, states appeared inactive) 31. IntelligenceEvolution: Sliding window learning rate - Was using lifetime cumulative average (never changed after early episodes). Now uses window of last 50 results 32. AdaptiveLearning: Normalized activityPattern - Was accumulating 0.1 per call (grew unbounded). Now stores raw counts, normalizes at read-time. feedbackHistory changed to ArrayDeque for O(1) removeFirst 33. CRZbrainMod: Fixed sender duplicate - allNearbyPlayers included the sender themselves. Added cleanup of lastChatLearnTime on player disconnect 34. ExternalAILearning: Removed double loadData() - init block was calling loadData() which was also called at server start 35. LCI: Debounced saveContextWeights - Added dirty flag, only saves when explicitly called instead of on every weight update

v2.2: Core AI Learning Infrastructure 1. Adaptive Learning Rate + Epsilon Decay - RL learning rate decreases over time, exploration decays (ReinforcementLearning.kt) 2. Experience Replay Buffer with Prioritized Sampling - RL replays high-reward episodes more often (ReinforcementLearning.kt) 3. Bilingual Sentiment IT/EN + Negation Detection - 3D sentiment (VAD) works in Italian and English, detects negation (AdvancedPatternMining.kt) 4. Quality Decay + Diversity Scoring - External AI responses lose quality over time, diverse responses preferred (ExternalAILearning.kt) 5. RL-Driven Response Selection - Response parts scored by RL success history (SmartResponseGenerator.kt) 6. Closed-Loop Feedback Attribution - Player feedback traces back to which response sources caused it (LearningContextIntegration.kt) 7. Cross-Pokemon Knowledge Transfer - Winning strategies propagate across trainers (GlobalKnowledgeNetwork.kt + AdaptiveLearning.kt) --- v2.3: Critical Bug Fixes + Exploration Overhaul 1. Move crash fix - null-safety on displayName.string, template access, target.pokemon (PokemonDefenseSystem.kt) 2. Quality score formula fix - Was cancelling out normalization (LearningContextIntegration.kt) 3. Policy Gradient fix - Now advantage-based (reward - baseline) instead of broken ln(probability) (ReinforcementLearning.kt) 4. UCB exploration - Replaces epsilon-greedy in both RL and AdaptiveLearning 5. Bayesian Beta confidence - Strategy success rates use proper Bayesian estimation (AdaptiveLearning.kt) 6. Strategy temporal decay - Old strategies gradually lose relevance (AdaptiveLearning.kt) 7. Markov chain prediction - Predicts trainer's next action from transition probabilities (AdaptiveLearning.kt) 8. Context weight persistence - Learned source weights saved/loaded as JSON (LearningContextIntegration.kt) 9. Conflict detection - Detects contradictions between context sources (LearningContextIntegration.kt) 10. Adaptive enhancement probabilities - contextQuality drives feature activation rates (LearningContextIntegration.kt) --- v2.4: RL Feedback Loop + Adaptive Probabilities 1. RL Feedback Loop closed - SRG records responses and detects player feedback (SmartResponseGenerator.kt) 2. LCI connected to SRG - Was completely dormant, now provides context (SmartResponseGenerator.kt) 3. Rich RL context - "event_nature_mood" instead of just "event_name" (SmartResponseGenerator.kt) 4. 6 adaptive gates - Key feature gates use getAdaptiveProbability() instead of hardcoded values (SmartResponseGenerator.kt) 5. RL exploration from profile - Uses learned epsilon instead of hardcoded 10% (SmartResponseGenerator.kt) 6. Context quality drives maxParts - 15-30% chance of 2 response parts based on quality (SmartResponseGenerator.kt) 7. RL scoring boost - Scaled by context quality, up to +25 points (SmartResponseGenerator.kt) 8. Post-processing with LCI context - Responses enhanced with learning context (SmartResponseGenerator.kt) --- v2.5: Terminology Learning 1. LearnedTerm data class - Tracks frequency, context, decay, anti-parrot flag (LearningSystem.kt) 2. learnPlayerTerminology() - Extracts words/expressions, filters IT/EN stop words (LearningSystem.kt) 3. getTopPlayerTerms() - Scored by frequency + recency - decay, minimum frequency 2 (LearningSystem.kt) 4. Terminology injection - ~20% chance in post-processing (SmartResponseGenerator.kt) 5. Nature-aware injection - jolly=excited, timid=hesitant, sassy=meta (SmartResponseGenerator.kt) 6. Multi-word expressions - 2-3 word phrases supported, cap 150 terms/player (LearningSystem.kt) 7. Persistence - LearningData v4, backward compatible with v3 (LearningSystem.kt) --- v2.6: Bug Fix + Dormant Code Activation 1. LanguageGrowth || → && fix - Expression/catchphrase development was 95% blocked (LanguageGrowth.kt) 2. LanguageGrowth typo + activation - Fixed "deplorevol mente", activated useAcquiredWord + generateEmotiveExpression (LanguageGrowth.kt) 3. EmotionalMomentum dynamics - Dynamic momentum by nature, energy/curiosity/confidence/stress influence responses (EmotionalMomentum.kt) 4. DeepUnderstanding 5x faster - Comprehension reaches 0.85 in 2K messages instead of 100K (DeepUnderstanding.kt) 5. IntelligenceEvolution inference fix - Fuzzy word matching for deductive, multi-word for inductive (IntelligenceEvolution.kt) 6. Battle dialogue probability - Nature/friendship-based: brave+max=67%, timid+min=7.5% (BattleEnhancementSystem.kt) 7. No early return in AdvancedIntelligence - Now flows through ALL post-processing systems (SmartResponseGenerator.kt) --- v2.7: TPS Optimization 1. PokemonDefenseSystem - Interval 5→20 ticks, 1 entity search per player instead of N*M, auto-deactivate (PokemonDefenseSystem.kt) 2. Async File I/O - MemorySystem.batchSave + RL + PatternMining + CollectiveIntelligence on CompletableFuture.runAsync (CRZbrainMod.kt) 3. Memory leak fix - 22 AI systems now have reset() and are registered in PokemonLifecycleManager 4. Chat learning throttle - 2s cooldown + max 3 Pokemon per message (was unlimited) (CRZbrainMod.kt) --- v2.8: Bug Fix + AI Quality Overhaul (44 fixes) Critical (10) 1. RL Experience Replay gradient fix - Was ln(prob), now advantage-based like main path (ReinforcementLearning.kt) 2. SRG Feedback Loop ordering - Read previous context BEFORE overwriting (was attributing to wrong context) (SmartResponseGenerator.kt) 3. LCI parser fix - Was parsing wrong format, always returning dummy data (LearningContextIntegration.kt) 4. LCI Quality Score fix - Removed division by qualityFactors (was ~0.17, now sum ~1.0) (LearningContextIntegration.kt) 5. EmbeddedAIProvider pokemonId fix - Pass UUID instead of display name (all lookups were failing) (EmbeddedAIProvider.kt) 6. ExternalAILearning usageCount - val→var + increment (overuse penalty was dead) (ExternalAILearning.kt) 7. ExternalAILearning 3 collections persisted - learnedByEvent/Emotion/Topic were lost on restart (ExternalAILearning.kt) 8. GlobalKnowledgeNetwork propagation - propagateWinningStrategy now actually calls AdaptiveLearning.importStrategy (GlobalKnowledgeNetwork.kt) 9. EmotionalMomentum loadAll() - Called at server start (states were saved but never loaded) (EmotionalMomentum.kt) 10. SRG confidence/curiosity passthrough - Were silently dropped when calling EmotionalMomentum.applyMomentum (SmartResponseGenerator.kt) High (14) 11. AdaptiveLearning temporal decay - Read-time only (was compounding destructively on every update) (AdaptiveLearning.kt) 12. RL double reward removed - integrateWithAdaptiveLearning no longer calls recordConsequence (ReinforcementLearning.kt) 13. RL UCB constant - Fixed to 1.414 (was decaying with epsilon, becoming useless) (ReinforcementLearning.kt) 14. RL plateau detection - Sliding window of last 200 rewards (was cumulative avg, always triggering) (ReinforcementLearning.kt) 15. LCI word-boundary matching - Feedback indicators use word boundaries (was substring: "sono"/"pokemon" triggered "no"/"ok") (LearningContextIntegration.kt) 16. APM transition matrix - Stores raw counts, normalizes on-the-fly (AdvancedPatternMining.kt) 17. LearningSystem action inference - Infers action type from message (question/greeting/farewell/compliment/command/chat) (LearningSystem.kt) 18. SRG wasSuccessful - Uses real feedback tracking (was always true) (SmartResponseGenerator.kt) 19. LanguageGrowth nature mapping - All 25 natures mapped to 4 categories + complexity 5 expressions (LanguageGrowth.kt) 20. SRG evolution keywords - Not filtered during EVOLUTION events (SmartResponseGenerator.kt) 21. ExternalAILearning separate .copy() - Per collection (feedback was multiplied 5x) (ExternalAILearning.kt) 22. AdaptiveLearning preferences - Non-destructive (both liked/disliked accumulate independently) (AdaptiveLearning.kt) 23. EmbeddedAIProvider level fix - {newLevel} off-by-one fixed (context.level is already the new level) (EmbeddedAIProvider.kt) 24. RL thread-safe replay - Local var instead of mutating shared profile.learningRate (ReinforcementLearning.kt) Medium + Optimizations (20) 25. RL prioritized sampling - Recalculates remaining priority after each selection 26. RL sliding window baseline - For advantage computation (replaces stale cumulative avg) 27. APM arousal baseline - 0.3 (was 0.5, wasting half the range) 28. APM English stopwords - 35 words added to keyword extraction 29. APM detectIntent - Word-boundary matching (was substring false-positive) 30. DeepUnderstanding random gate - On output not analysis (detectImplicit, detectSarcasm) 31. DeepUnderstanding emotionalReading - No longer multiplies intensity (was making emotion path unreachable) 32. IntelligenceEvolution inference - Single-roll weighted type selection (was biased cascade) 33. EmotionalMomentum punctuation - Regex-based period/exclamation replacement (sentence-ending only) 34. ExternalAILearning recencyBonus - Uses creation timestamp (was rich-get-richer on lastUsed) 35. ExternalAILearning Gson fixup - Post-load fixup for pre-v2.2 data 36. GlobalKnowledgeNetwork diversity - Uses total trainers (was always 1.0) 37. SRG thread-safe maps - ConcurrentHashMap for lastEventByPokemon/recentEventsByPokemon 38. SRG reset() - Cleans all maps (was missing 2+1 maps) 39. LanguageGrowth faster progression - Lower divisors + Math.round 40. SRG discard reduced - ContextualResponse 40%→15% 41. SRG bounded action space - Uses source category instead of response text 42. APM sentiment cache - LRU eviction when full (was stopping caching at 500) 43. EmbeddedAIProvider reset() - Added + registered in PokemonLifecycleManager (4 maps cleaned) 44. EmotionalMomentum reset() - Deletes JSON file (was leaving orphans on disk) --- v2.9: Performance Optimization 1. LCI containsWord→Set check - Split once, Set.contains(), zero Regex per call (was 60+) (LearningContextIntegration.kt) 2. LCI 4 parser Regex pre-compiled - PARSER_MESSAGE_COUNT, PARSER_MOOD, etc. (LearningContextIntegration.kt) 3. APM detectIntent rewrite - Set-based word matching (was 20+ Regex per call) (AdvancedPatternMining.kt) 4. APM sentiment cache LRU - Eviction when full (was stopping caching) (AdvancedPatternMining.kt) 5. SRG wordBoundary→Set check - Was 24 Regex per call (SmartResponseGenerator.kt) 6. SRG removed duplicate learning - Was running 17 ops twice per response (SmartResponseGenerator.kt) 7. SmartConversation simplified - Replaced with recordDirectInteraction only (SmartConversation.kt) 8. DeepUnderstanding 36 Regex pre-compiled - Object-level vals (was per-message creation) (DeepUnderstanding.kt) 9. LearningSystem VOCAB_SPLIT_REGEX - Pre-compiled, recordDirectInteraction made public (LearningSystem.kt) 10. CRZbrainMod isDirectToMe - true for sender's own Pokemon in handlePlayerChat (CRZbrainMod.kt) 11. LearningSystem size caps - learnedVocabulary(500), associations(300), behaviorPatterns(100), positiveResponseTriggers(50), evidence(20) (LearningSystem.kt) Impact: ~150 Regex/message → ~5, learning ops halved for responding Pokemon --- v3.0: AI Quality Overhaul (Zero CPU Cost) Gruppo A: Emotional + Intention Scoring 1. EmotionalMomentum guides selection - Emotional state influences WHICH response is chosen, not just post-processing (SmartResponseGenerator.kt) 2. DeepUnderstanding moved before selection - analyzeMessage runs BEFORE scoring, intent guides response choice (SmartResponseGenerator.kt) 3. Intention-based scoring - QUESTION→answers, PRAISE→thanks, CRITICISM→apologies, AFFECTION→emotes (SmartResponseGenerator.kt) Gruppo B: Waste Reduction 4. Score-based rejection - topScore>=20 always passes, else 75% (was 50% random discard) (SmartResponseGenerator.kt) 5. RL boost 2.5x stronger - Formula (40+ctx20) range 0-60 (was 20+ctx5, range 0-25) (SmartResponseGenerator.kt) 6. IQ-based comment chance - 15-35% based on IQ (was fixed 15%) (SmartResponseGenerator.kt) 7. Pattern-based recall chance - 10-25% based on patternRecognition (was fixed 10%) (SmartResponseGenerator.kt) Gruppo C: Adaptive Gates 8. 7 more gates adaptive - species_behavior, social_comment, battle_style, proactive_comment, help_offer, rank_dialogue, inter_member (SmartResponseGenerator.kt) 9. Terminology + DeepUnderstanding adaptive - Both injection gates use getAdaptiveProbability() (SmartResponseGenerator.kt) Gruppo D: Transitions + Feedback Loop 10. Content-based transitions - combineResponsesLogically uses content analysis instead of random (SmartResponseGenerator.kt) 11. Feature source tracking - activeFeatureSources feeds into recordResponseAttribution for feedback (SmartResponseGenerator.kt) Gruppo E: Complexity-Aware Depth 12. getComplexityLevel() added - Public function in LanguageGrowth (LanguageGrowth.kt) 13. High-complexity Pokemon depth - complexity>=4 with high abstractThinking get extra intelligent comments (SmartResponseGenerator.kt) --- v3.1: Stability + Performance + Dormant AI Activation Gruppo A: Stability/Persistence (6 fixes) 1. Shutdown async wait - Server waits up to 5s for async save to finish before shutdown saves (CRZbrainMod.kt) 2. ExternalAILearning.loadData() at startup - Was missing, 3 collections not loaded (CRZbrainMod.kt) 3. Atomic write - LearningSystem writes to .tmp then renames, prevents corruption on crash (LearningSystem.kt) 4. isDirty timing verified - CollectiveIntelligence was already correct, added clarifying comments (CollectiveIntelligence.kt) 5. learnedClosers persisted - Now saved/loaded as learned_closers.json (was lost every restart) (ExternalAILearning.kt) 6. recentEpisodeRewards verified - Already persisted via Gson in RLProfile data class (ReinforcementLearning.kt) Gruppo B: Performance (13 fixes) 7. IntelligenceEvolution 17 Regex pre-compiled - detectEmotionalPattern(5), detectRequestPattern(4), categorizeMessage(8) (IntelligenceEvolution.kt) 8. 4x removeAt(0) → ArrayDeque.removeFirst() - O(1) instead of O(n) in SRG, APM, EmbeddedAIProvider(2x) 9. Outer map cap 200 - EmbeddedAIProvider trimOuterMapsIfNeeded() prevents unbounded growth (EmbeddedAIProvider.kt) 10. SRG CLEAN_ACTION_REGEX - Pre-compiled Regex in scoring loop (SmartResponseGenerator.kt) 11. AIHandler 2 retry Regex - CLAUDE_RETRY_REGEX + RETRY_DELAY_REGEX pre-compiled (AIHandler.kt) 12. ContextualMemoryBridge 2 Regex - DIGIT_REGEX + INTENSITY_REGEX pre-compiled (ContextualMemoryBridge.kt) 13. ExternalAILearning 3 Regex - SENTENCE_SPLIT_REGEX + WORD_SPLIT_REGEX + ASTERISK_EXPRESSION_REGEX pre-compiled (ExternalAILearning.kt) 14. AdvancedPersonalitySystem 1 Regex - TOKEN_SPLIT_REGEX pre-compiled (AdvancedPersonalitySystem.kt) Gruppo C: Dormant AI Activation (3 fixes) 15. predictNextAction activated - Markov chain predictions enrich RL context before response selection (SmartResponseGenerator.kt ← AdvancedPatternMining.kt) 16. validatePrediction activated - Predictions validated on next player message, closes feedback loop (SmartResponseGenerator.kt ← AdvancedPatternMining.kt) 17. Cross-Pokemon RL transfer - New Pokemon (<10 episodes) get +15 score boost from global top actions; integrateWithAdaptiveLearning syncs every 50 episodes (SmartResponseGenerator.kt ← ReinforcementLearning.kt)

● v2.7 TPS Optimization - Summary Fix 1: PokemonDefenseSystem (HIGH impact) - Tick interval increased from 5 (0.25s) to 20 ticks (1s) → 75% fewer calls - Entity search now runs once per player instead of once per Pokemon per player (was N*M, now 1*N). If 10 players have 6 Pokemon each, that's 60 AABB searches → 10 - Auto-deactivate: defense mode automatically disables when no threats remain, preventing useless detectThreats() calls every tick Fix 2: Async File I/O (HIGH impact) - MemorySystem.batchSave(), ReinforcementLearning.saveData(), AdvancedPatternMining.saveData(), and CollectiveIntelligence.saveData() now all run on CompletableFuture.runAsync instead of the main server thread - AtomicBoolean lock prevents concurrent saves from overlapping - Main thread no longer freezes during JSON serialization (was several MB every 60s) Fix 3: Memory Leak Prevention (22 systems, MEDIUM impact) - Added reset(pokemonId) functions to 19 AI systems that had none (EmotionalMomentum, ConversationMemory, NarrativeEngine, DeepThinking, DeepPersonality, UltraConsciousness, ProactiveAI, VoicePatterns, LexicalVariety, AdvancedPersonalitySystem, MasterMind, EmotionalIntelligence, PokemonSocialAI, SmartResponseGenerator, PokemonGoals, DynamicPersonality, PokemonLifeStory, PokemonBondSystem, LearningContextIntegration) - Wired 4 existing resets that were never connected (PersonalityEvolution, MoodSystem, SmartConversation, WisdomSystem) - PokemonLifecycleManager now cleans up all 39 systems on stale Pokemon eviction (was only 17) Fix 4: Chat Learning Throttle (MEDIUM impact) - 2-second cooldown between learning triggers from the same player - Max 3 Pokemon learn per chat message (was unlimited - every Pokemon of every nearby player) - Drastically reduces CPU load on servers with active chat Estimated TPS improvement ┌─────────┬─────────────────┬──────────────┐ │ Players │ Before │ After │ ├─────────┼─────────────────┼──────────────┤ │ 1-2 │ ~0.5-1 TPS lost │ ~0.1-0.2 TPS │ ├─────────┼─────────────────┼──────────────┤ │ 5-10 │ ~1-3 TPS lost │ ~0.3-0.7 TPS │ ├─────────┼─────────────────┼──────────────┤ │ 20+ │ ~3-5+ TPS lost │ ~0.5-1.5 TPS │ └─────────┴─────────────────┴──────────────┘

Bug Fix: Confused conversations when multiple Pokemon are out Problem: When multiple Pokemon were out of their Pokeballs at the same time, they would say things that didn't make sense - wrong personality, wrong stats, mixed-up context. Root causes found and fixed: Fix 1 — ResponseContext was shared (AIHandler.kt) - Before: A single ResponseContext was built using only the first Pokemon's data (level, health, friendship, mood). All Pokemon used that same context. - After: Added buildContextForPokemon() that creates a separate context per Pokemon with its own level, HP, friendship, and mood. Fix 2 — Cache keys used name instead of UUID (EmbeddedAIProvider.kt) - Before: Response caches (recentResponses, recentStructures, responseCounters) were keyed by Pokemon display name. Two Pokemon with the same species/nickname would share caches, corrupting response variety and adaptive probabilities. Also ConversationMemory was called with the name instead of UUID, so contextual responses never worked. - After: Added pokemonId parameter. All caches now use UUID as key. No more collisions between same-name Pokemon. Fix 3 — DialogueSystem matched wrong Pokemon (DialogueSystem.kt) - Before: pokemons.find() always returned the first match by name. With two Pokemon of the same species, the second one never got its speech bubble. - After: Added matchedPokemonUuids tracking set. Each dialogue is assigned to a different Pokemon entity, even if they share the same name. Fix 4 — MultiPokemonChatSystem was Italian-only - Before: All interaction responses (agree, disagree, joke, question, addition) and message detection (addressed to all, discussion topics, addressed Pokemon) were Italian-only. - After: All 8 functions now use Lang.tList() for bilingual IT/EN support. Bonus fixes - Lang.kt — Made t() and tList() generic so they work with data classes (Fear, Secret, Belief, Legend), fixing pre-existing compile errors in PokemonLifeStory.kt and PokemonLore.kt. - MoodSystem.kt — 16 mood descriptions + getMoodDescription() now bilingual. - WorldContext.kt — toPromptText(), toCompactText(), and collect() (weather, time, biome) now bilingual.

CRZbrain is a server-side Cobblemon plugin (Fabric mod for Minecraft 1.7.1) that adds an intelligent AI personality system to Pokemon. Here's what it does: - Smart Responses: Each Pokemon can generate context-aware responses using an embedded AI engine (SmartResponseGenerator), reacting to battles, interactions, and trainer behavior with unique personality - Reinforcement Learning: Pokemon learn from experience over time — successful interactions are reinforced, failed ones are penalized, using policy gradient methods with adaptive learning rates and experience replay (ReinforcementLearning.kt) - Sentiment Analysis: The system understands player messages in both Italian and English, analyzing emotional valence, arousal, and dominance (3D VAD model) with negation detection (AdvancedPatternMining.kt) - Adaptive Learning: Pokemon track what works and what doesn't, balancing exploration of new behaviors vs exploitation of proven ones (AdaptiveLearning.kt) - External AI Integration: Can optionally learn from external AI providers like Claude, OpenAI, and Gemini to enrich response quality (ExternalAILearning.kt) - Global Knowledge Network: Pokemon share knowledge across trainers — if one trainer teaches something, other Pokemon can benefit through consensus-based knowledge transfer (GlobalKnowledgeNetwork.kt) - Behavior Prediction & Topic Modeling: The system mines patterns from interactions to predict player behavior and understand conversation topics (AdvancedPatternMining.kt) In short: it makes Pokemon feel alive — they learn, adapt, remember, and grow smarter the more you interact with them.