[upd] Implement BDI reasoning system and enhance AI decision-making with persistent beliefs and desires. Introduced Redis state storage for improved simulation state management and added asynchronous logging for performance optimization. Updated configuration parameters to support new features.

backend/config.py

@@ -133,7 +133,7 @@ class EconomyConfig:
 
 @dataclass
 class AIConfig:
-    """Configuration for AI decision-making system (GOAP-based)."""
+    """Configuration for AI decision-making system."""
     # Maximum A* iterations for GOAP planner
     goap_max_iterations: int = 50
 
@@ -143,6 +143,53 @@ class AIConfig:
     # Fall back to reactive planning if GOAP fails to find a plan
     reactive_fallback: bool = True
 
+    # Use BDI (Belief-Desire-Intention) instead of pure GOAP
+    # BDI adds persistent beliefs, long-term desires, and plan commitment
+    use_bdi: bool = False
+
+
+@dataclass
+class BDIConfig:
+    """Configuration for BDI (Belief-Desire-Intention) reasoning system.
+    
+    BDI extends GOAP with:
+    - Persistent beliefs (memory of past events)
+    - Long-term desires (personality-driven motivations)
+    - Committed intentions (plan persistence)
+    """
+    # Timeslicing: how often agents run full deliberation
+    # 1 = every turn, 3 = every 3rd turn (staggered by agent ID)
+    thinking_interval: int = 1
+
+    # Maximum consecutive action failures before replanning
+    max_consecutive_failures: int = 2
+
+    # Priority multiplier needed to switch from current intention
+    # 1.5 = new goal must be 50% higher priority to cause a switch
+    priority_switch_threshold: float = 1.5
+
+    # Memory system settings
+    memory_max_events: int = 50      # Max events to remember
+    memory_decay_rate: float = 0.1   # How fast memories fade
+
+
+@dataclass
+class RedisConfig:
+    """Configuration for optional Redis state storage.
+    
+    Redis enables:
+    - Persistent state across restarts
+    - Decoupled UI polling (web clients read independently)
+    - Distributed access (multiple simulation instances)
+    """
+    enabled: bool = False
+    host: str = "localhost"
+    port: int = 6379
+    db: int = 0
+    password: Optional[str] = None
+    prefix: str = "villsim:"
+    ttl_seconds: int = 3600  # 1 hour default TTL
+
 
 @dataclass
 class AgeConfig:
@@ -252,6 +299,7 @@ class PerformanceConfig:
     # Logging control
     logging_enabled: bool = False           # Enable file logging (disable for speed)
     detailed_logging: bool = False          # Enable verbose per-agent logging
+    async_logging: bool = True              # Use non-blocking background logging
     log_flush_interval: int = 50            # Flush logs every N turns (not every turn)
 
     # Memory management
@@ -260,6 +308,9 @@ class PerformanceConfig:
     # Statistics calculation frequency
     stats_update_interval: int = 10         # Update expensive stats every N turns
 
+    # State storage
+    state_storage_enabled: bool = True      # Enable state snapshotting
+
 
 @dataclass
 class SimulationConfig:
@@ -272,6 +323,8 @@ class SimulationConfig:
     market: MarketConfig = field(default_factory=MarketConfig)
     economy: EconomyConfig = field(default_factory=EconomyConfig)
     ai: AIConfig = field(default_factory=AIConfig)
+    bdi: BDIConfig = field(default_factory=BDIConfig)
+    redis: RedisConfig = field(default_factory=RedisConfig)
     age: AgeConfig = field(default_factory=AgeConfig)
     storage: StorageConfig = field(default_factory=StorageConfig)
     sinks: SinksConfig = field(default_factory=SinksConfig)
@@ -281,9 +334,10 @@ class SimulationConfig:
 
     def to_dict(self) -> dict:
         """Convert to dictionary."""
-        return {
+        result = {
             "performance": asdict(self.performance),
             "ai": asdict(self.ai),
+            "bdi": asdict(self.bdi),
             "agent_stats": asdict(self.agent_stats),
             "resources": asdict(self.resources),
             "actions": asdict(self.actions),
@@ -295,13 +349,24 @@ class SimulationConfig:
             "sinks": asdict(self.sinks),
             "auto_step_interval": self.auto_step_interval,
         }
+        # Handle redis separately due to Optional field
+        redis_dict = asdict(self.redis)
+        result["redis"] = redis_dict
+        return result
 
     @classmethod
     def from_dict(cls, data: dict) -> "SimulationConfig":
         """Create from dictionary."""
+        # Handle redis config specially due to Optional password
+        redis_data = data.get("redis", {})
+        if redis_data.get("password") is None:
+            redis_data["password"] = None
+
         return cls(
             performance=PerformanceConfig(**data.get("performance", {})),
             ai=AIConfig(**data.get("ai", {})),
+            bdi=BDIConfig(**data.get("bdi", {})),
+            redis=RedisConfig(**redis_data),
             agent_stats=AgentStatsConfig(**data.get("agent_stats", {})),
             resources=ResourceConfig(**data.get("resources", {})),
             actions=ActionConfig(**data.get("actions", {})),
@@ -407,3 +472,8 @@ def _reset_all_caches() -> None:
     except ImportError:
         pass
 
+    try:
+        from backend.core.storage import reset_state_store
+        reset_state_store()
+    except ImportError:
+        pass

backend/core/ai.py

@@ -1,13 +1,14 @@
 """AI decision system for agents in the Village Simulation.
 
-This module provides shared AI classes and utilities used by the GOAP
+This module provides the main entry point for AI decisions, supporting
-(Goal-Oriented Action Planning) system.
+both GOAP (Goal-Oriented Action Planning) and BDI (Belief-Desire-Intention)
+reasoning systems.
 
-GOAP Benefits:
+AI System Options:
-- Agents plan multi-step sequences to achieve goals
+- GOAP: Fast, reactive planning with goal prioritization
-- Goals are dynamically prioritized based on state
+- BDI: Persistent beliefs, long-term desires, plan commitment
-- More emergent and adaptive behavior
+
-- Easier to extend with new goals and actions
+Configure via config.json "ai.use_bdi" (default: false for backward compatibility)
 
 Major features:
 - Each agent has unique personality traits affecting all decisions
@@ -51,6 +52,10 @@ class AIDecision:
     # For price adjustments
     adjust_order_id: Optional[str] = None
     new_price: Optional[int] = None
+    # GOAP/BDI fields
+    goal_name: str = ""
+    plan_length: int = 0
+    bdi_info: dict = field(default_factory=dict)
 
     def to_dict(self) -> dict:
         return {
@@ -71,6 +76,9 @@ class AIDecision:
             ],
             "adjust_order_id": self.adjust_order_id,
             "new_price": self.new_price,
+            "goal_name": self.goal_name,
+            "plan_length": self.plan_length,
+            "bdi_info": self.bdi_info,
         }
 
 
@@ -101,6 +109,7 @@ def get_energy_cost(resource_type: ResourceType) -> int:
 # Cached config values to avoid repeated lookups
 _cached_ai_config = None
 _cached_economy_config = None
+_cached_use_bdi = None
 
 
 def _get_ai_config():
@@ -121,11 +130,30 @@ def _get_economy_config():
     return _cached_economy_config
 
 
+def _should_use_bdi() -> bool:
+    """Check if BDI should be used (cached)."""
+    global _cached_use_bdi
+    if _cached_use_bdi is None:
+        from backend.config import get_config
+        config = get_config()
+        ai_config = getattr(config, 'ai', None)
+        _cached_use_bdi = getattr(ai_config, 'use_bdi', False) if ai_config else False
+    return _cached_use_bdi
+
+
 def reset_ai_config_cache():
     """Reset the cached config values (call after config reload)."""
-    global _cached_ai_config, _cached_economy_config
+    global _cached_ai_config, _cached_economy_config, _cached_use_bdi
     _cached_ai_config = None
     _cached_economy_config = None
+    _cached_use_bdi = None
+    
+    # Also reset BDI state if it was being used
+    try:
+        from backend.core.bdi import reset_bdi_state
+        reset_bdi_state()
+    except ImportError:
+        pass
 
 
 def get_ai_decision(
@@ -136,7 +164,9 @@ def get_ai_decision(
     current_turn: int = 0,
     is_night: bool = False,
 ) -> AIDecision:
-    """Get an AI decision for an agent using GOAP (Goal-Oriented Action Planning).
+    """Get an AI decision for an agent.
+    
+    Uses either GOAP or BDI based on config setting "ai.use_bdi".
     
     Args:
         agent: The agent to make a decision for
@@ -149,8 +179,38 @@ def get_ai_decision(
     Returns:
         AIDecision with the chosen action and parameters
     """
+    if _should_use_bdi():
+        return _get_bdi_decision(
+            agent=agent,
+            market=market,
+            step_in_day=step_in_day,
+            day_steps=day_steps,
+            current_turn=current_turn,
+            is_night=is_night,
+        )
+    else:
+        return _get_goap_decision(
+            agent=agent,
+            market=market,
+            step_in_day=step_in_day,
+            day_steps=day_steps,
+            current_turn=current_turn,
+            is_night=is_night,
+        )
+
+
+def _get_goap_decision(
+    agent: Agent,
+    market: "OrderBook",
+    step_in_day: int,
+    day_steps: int,
+    current_turn: int,
+    is_night: bool,
+) -> AIDecision:
+    """Get an AI decision using GOAP (Goal-Oriented Action Planning)."""
     from backend.core.goap.goap_ai import get_goap_decision
-    return get_goap_decision(
+
+    goap_decision = get_goap_decision(
         agent=agent,
         market=market,
         step_in_day=step_in_day,
@@ -158,3 +218,84 @@ def get_ai_decision(
         current_turn=current_turn,
         is_night=is_night,
     )
+    
+    # Convert GOAP AIDecision to our AIDecision (they should be compatible)
+    return AIDecision(
+        action=goap_decision.action,
+        target_resource=goap_decision.target_resource,
+        order_id=goap_decision.order_id,
+        quantity=goap_decision.quantity,
+        price=goap_decision.price,
+        reason=goap_decision.reason,
+        trade_items=[
+            TradeItem(
+                order_id=t.order_id,
+                resource_type=t.resource_type,
+                quantity=t.quantity,
+                price_per_unit=t.price_per_unit,
+            )
+            for t in goap_decision.trade_items
+        ],
+        adjust_order_id=goap_decision.adjust_order_id,
+        new_price=goap_decision.new_price,
+        goal_name=goap_decision.goal_name,
+        plan_length=goap_decision.plan_length,
+    )
+
+
+def _get_bdi_decision(
+    agent: Agent,
+    market: "OrderBook",
+    step_in_day: int,
+    day_steps: int,
+    current_turn: int,
+    is_night: bool,
+) -> AIDecision:
+    """Get an AI decision using BDI (Belief-Desire-Intention)."""
+    from backend.core.bdi import get_bdi_decision
+
+    bdi_decision = get_bdi_decision(
+        agent=agent,
+        market=market,
+        step_in_day=step_in_day,
+        day_steps=day_steps,
+        current_turn=current_turn,
+        is_night=is_night,
+    )
+    
+    # Convert BDI AIDecision to our AIDecision
+    return AIDecision(
+        action=bdi_decision.action,
+        target_resource=bdi_decision.target_resource,
+        order_id=bdi_decision.order_id,
+        quantity=bdi_decision.quantity,
+        price=bdi_decision.price,
+        reason=bdi_decision.reason,
+        trade_items=[
+            TradeItem(
+                order_id=t.order_id,
+                resource_type=t.resource_type,
+                quantity=t.quantity,
+                price_per_unit=t.price_per_unit,
+            )
+            for t in bdi_decision.trade_items
+        ],
+        adjust_order_id=bdi_decision.adjust_order_id,
+        new_price=bdi_decision.new_price,
+        goal_name=bdi_decision.goal_name,
+        plan_length=bdi_decision.plan_length,
+        bdi_info=bdi_decision.bdi_info,
+    )
+
+
+def on_agent_death(agent_id: str) -> None:
+    """Clean up AI state when an agent dies.
+    
+    Call this from the engine when an agent is removed.
+    """
+    if _should_use_bdi():
+        try:
+            from backend.core.bdi import remove_agent_bdi_state
+            remove_agent_bdi_state(agent_id)
+        except ImportError:
+            pass

backend/core/bdi/__init__.py

@@ -0,0 +1,50 @@
+"""BDI (Belief-Desire-Intention) module for agent AI.
+
+This module provides a BDI architecture that wraps the existing GOAP planner,
+enabling:
+- Persistent beliefs (memory of past events)
+- Long-term desires (personality-driven motivations)
+- Committed intentions (plan persistence)
+
+Main entry points:
+- get_bdi_decision(): Get an AI decision using BDI reasoning
+- reset_bdi_state(): Reset all agent BDI state (on simulation reset)
+"""
+
+from backend.core.bdi.belief import BeliefBase, MemoryEvent
+from backend.core.bdi.desire import Desire, DesireType, DesireManager
+from backend.core.bdi.intention import (
+    Intention,
+    IntentionManager,
+    CommitmentStrategy,
+)
+from backend.core.bdi.bdi_agent import (
+    BDIAgentAI,
+    AIDecision,
+    TradeItem,
+    get_bdi_decision,
+    reset_bdi_state,
+    remove_agent_bdi_state,
+)
+
+__all__ = [
+    # Belief system
+    "BeliefBase",
+    "MemoryEvent",
+    # Desire system
+    "Desire",
+    "DesireType",
+    "DesireManager",
+    # Intention system
+    "Intention",
+    "IntentionManager",
+    "CommitmentStrategy",
+    # BDI Agent
+    "BDIAgentAI",
+    "AIDecision",
+    "TradeItem",
+    # Entry points
+    "get_bdi_decision",
+    "reset_bdi_state",
+    "remove_agent_bdi_state",
+]

backend/core/bdi/bdi_agent.py

@@ -0,0 +1,627 @@
+"""BDI Agent AI that wraps GOAP planning with BDI reasoning.
+
+This module provides the main BDI-based AI decision maker that:
+1. Maintains persistent beliefs about the world
+2. Manages desires based on personality
+3. Commits to intentions (plans) and executes them
+4. Uses GOAP planning to generate action sequences
+
+Performance optimizations:
+- Timeslicing: full BDI cycle only runs periodically
+- Plan persistence: reuses plans across turns
+- Cached belief updates: skips unchanged data
+"""
+
+from dataclasses import dataclass, field
+from typing import Optional, TYPE_CHECKING
+
+from backend.domain.action import ActionType
+from backend.domain.resources import ResourceType
+from backend.domain.personality import get_trade_price_modifier
+
+from backend.core.bdi.belief import BeliefBase
+from backend.core.bdi.desire import DesireManager
+from backend.core.bdi.intention import IntentionManager
+
+from backend.core.goap.planner import GOAPPlanner, ReactivePlanner
+from backend.core.goap.goals import get_all_goals
+from backend.core.goap.actions import get_all_actions
+
+if TYPE_CHECKING:
+    from backend.domain.agent import Agent
+    from backend.core.market import OrderBook
+    from backend.core.goap.goal import Goal
+    from backend.core.goap.action import GOAPAction
+    from backend.core.goap.planner import Plan
+
+
+@dataclass
+class TradeItem:
+    """A single item to buy/sell in a trade."""
+    order_id: str
+    resource_type: ResourceType
+    quantity: int
+    price_per_unit: int
+
+
+@dataclass
+class AIDecision:
+    """A decision made by the AI for an agent."""
+    action: ActionType
+    target_resource: Optional[ResourceType] = None
+    order_id: Optional[str] = None
+    quantity: int = 1
+    price: int = 0
+    reason: str = ""
+    trade_items: list[TradeItem] = field(default_factory=list)
+    adjust_order_id: Optional[str] = None
+    new_price: Optional[int] = None
+
+    # GOAP/BDI-specific fields
+    goal_name: str = ""
+    plan_length: int = 0
+    bdi_info: dict = field(default_factory=dict)
+
+    def to_dict(self) -> dict:
+        return {
+            "action": self.action.value,
+            "target_resource": self.target_resource.value if self.target_resource else None,
+            "order_id": self.order_id,
+            "quantity": self.quantity,
+            "price": self.price,
+            "reason": self.reason,
+            "trade_items": [
+                {
+                    "order_id": t.order_id,
+                    "resource_type": t.resource_type.value,
+                    "quantity": t.quantity,
+                    "price_per_unit": t.price_per_unit,
+                }
+                for t in self.trade_items
+            ],
+            "adjust_order_id": self.adjust_order_id,
+            "new_price": self.new_price,
+            "goal_name": self.goal_name,
+            "plan_length": self.plan_length,
+            "bdi_info": self.bdi_info,
+        }
+
+
+class BDIAgentAI:
+    """BDI-based AI decision maker that wraps GOAP planning.
+    
+    The BDI cycle:
+    1. Update beliefs from sensors (agent state, market)
+    2. Update desires based on beliefs and personality
+    3. Check if current intention should continue
+    4. If needed, generate new plan via GOAP
+    5. Execute next action from intention
+    
+    Performance features:
+    - Timeslicing: full deliberation only every N turns
+    - Plan persistence: reuse plans across turns
+    - Reactive fallback: simple decisions when not deliberating
+    """
+    
+    # Class-level cache for planners (shared across instances)
+    _planner_cache: Optional[GOAPPlanner] = None
+    _reactive_cache: Optional[ReactivePlanner] = None
+    _goals_cache: Optional[list] = None
+    _actions_cache: Optional[list] = None
+    
+    def __init__(
+        self,
+        agent: "Agent",
+        market: "OrderBook",
+        step_in_day: int = 1,
+        day_steps: int = 10,
+        current_turn: int = 0,
+        is_night: bool = False,
+        # Persistent BDI state (passed in for continuity)
+        beliefs: Optional[BeliefBase] = None,
+        desires: Optional[DesireManager] = None,
+        intentions: Optional[IntentionManager] = None,
+    ):
+        self.agent = agent
+        self.market = market
+        self.step_in_day = step_in_day
+        self.day_steps = day_steps
+        self.current_turn = current_turn
+        self.is_night = is_night
+        
+        # Initialize or use existing BDI components
+        self.beliefs = beliefs or BeliefBase()
+        self.desires = desires or DesireManager(agent.personality)
+        self.intentions = intentions or IntentionManager.from_personality(agent.personality)
+        
+        # Update beliefs from current state
+        self.beliefs.update_from_sensors(
+            agent=agent,
+            market=market,
+            step_in_day=step_in_day,
+            day_steps=day_steps,
+            current_turn=current_turn,
+            is_night=is_night,
+        )
+        
+        # Update desires from beliefs
+        self.desires.update_from_beliefs(self.beliefs)
+        
+        # Get cached planners and goals/actions
+        self.planner = self._get_planner()
+        self.reactive_planner = self._get_reactive_planner()
+        self.goals = self._get_goals()
+        self.actions = self._get_actions()
+        
+        # Personality shortcuts
+        self.p = agent.personality
+        self.skills = agent.skills
+    
+    @classmethod
+    def _get_planner(cls) -> GOAPPlanner:
+        """Get cached GOAP planner."""
+        if cls._planner_cache is None:
+            from backend.config import get_config
+            config = get_config()
+            ai_config = config.ai
+            cls._planner_cache = GOAPPlanner(
+                max_iterations=ai_config.goap_max_iterations,
+            )
+        return cls._planner_cache
+    
+    @classmethod
+    def _get_reactive_planner(cls) -> ReactivePlanner:
+        """Get cached reactive planner."""
+        if cls._reactive_cache is None:
+            cls._reactive_cache = ReactivePlanner()
+        return cls._reactive_cache
+    
+    @classmethod
+    def _get_goals(cls) -> list:
+        """Get cached goals list."""
+        if cls._goals_cache is None:
+            cls._goals_cache = get_all_goals()
+        return cls._goals_cache
+    
+    @classmethod
+    def _get_actions(cls) -> list:
+        """Get cached actions list."""
+        if cls._actions_cache is None:
+            cls._actions_cache = get_all_actions()
+        return cls._actions_cache
+    
+    @classmethod
+    def reset_caches(cls) -> None:
+        """Reset all caches (call after config reload)."""
+        cls._planner_cache = None
+        cls._reactive_cache = None
+        cls._goals_cache = None
+        cls._actions_cache = None
+    
+    def should_deliberate(self) -> bool:
+        """Check if this agent should run full BDI deliberation this turn.
+        
+        Timeslicing: not every agent deliberates every turn.
+        Agents are staggered based on their ID hash.
+        """
+        from backend.config import get_config
+        config = get_config()
+        
+        # Get thinking interval from config (default to 1 = every turn)
+        bdi_config = getattr(config, 'bdi', None)
+        thinking_interval = getattr(bdi_config, 'thinking_interval', 1) if bdi_config else 1
+        
+        if thinking_interval <= 1:
+            return True  # Deliberate every turn
+        
+        # Stagger agents across turns
+        agent_hash = hash(self.agent.id) % thinking_interval
+        return (self.current_turn % thinking_interval) == agent_hash
+    
+    def decide(self) -> AIDecision:
+        """Make a decision using BDI reasoning with GOAP planning.
+        
+        Decision flow:
+        1. Night time: mandatory sleep
+        2. Check if should deliberate (timeslicing)
+        3. If deliberating: run full BDI cycle
+        4. If not: continue current intention or reactive fallback
+        """
+        # Night time - mandatory sleep
+        if self.is_night:
+            return AIDecision(
+                action=ActionType.SLEEP,
+                reason="Night time: sleeping",
+                goal_name="Sleep",
+                bdi_info={"mode": "night"},
+            )
+        
+        # Check if we should run full deliberation
+        if self.should_deliberate():
+            return self._deliberate()
+        else:
+            return self._continue_or_react()
+    
+    def _deliberate(self) -> AIDecision:
+        """Run full BDI deliberation cycle."""
+        # Filter goals by desires
+        filtered_goals = self.desires.filter_goals_by_desire(self.goals, self.beliefs)
+        
+        # Check if we should reconsider current intention
+        should_replan = self.intentions.should_reconsider(
+            beliefs=self.beliefs,
+            desire_manager=self.desires,
+            available_goals=filtered_goals,
+        )
+        
+        if not should_replan and self.intentions.has_intention():
+            # Continue with current intention
+            action = self.intentions.get_next_action()
+            if action:
+                return self._convert_to_decision(
+                    goap_action=action,
+                    goal=self.intentions.current_intention.goal,
+                    plan=self.intentions.current_intention.plan,
+                    mode="continue",
+                )
+        
+        # Need to plan for a goal
+        world_state = self.beliefs.to_world_state()
+        
+        plan = self.planner.plan_for_goals(
+            initial_state=world_state,
+            goals=filtered_goals,
+            available_actions=self.actions,
+        )
+        
+        if plan and not plan.is_empty:
+            # Commit to new intention
+            self.intentions.commit_to_plan(
+                goal=plan.goal,
+                plan=plan,
+                current_turn=self.current_turn,
+            )
+            
+            goap_action = plan.first_action
+            return self._convert_to_decision(
+                goap_action=goap_action,
+                goal=plan.goal,
+                plan=plan,
+                mode="new_plan",
+            )
+        
+        # Fallback to reactive planning
+        return self._reactive_fallback()
+    
+    def _continue_or_react(self) -> AIDecision:
+        """Continue current intention or use reactive fallback (no deliberation)."""
+        if self.intentions.has_intention():
+            action = self.intentions.get_next_action()
+            if action:
+                return self._convert_to_decision(
+                    goap_action=action,
+                    goal=self.intentions.current_intention.goal,
+                    plan=self.intentions.current_intention.plan,
+                    mode="timeslice_continue",
+                )
+        
+        # No intention, use reactive fallback
+        return self._reactive_fallback()
+    
+    def _reactive_fallback(self) -> AIDecision:
+        """Use reactive planning when no intention exists."""
+        world_state = self.beliefs.to_world_state()
+        
+        best_action = self.reactive_planner.select_best_action(
+            state=world_state,
+            goals=self.goals,
+            available_actions=self.actions,
+        )
+        
+        if best_action:
+            return self._convert_to_decision(
+                goap_action=best_action,
+                goal=None,
+                plan=None,
+                mode="reactive",
+            )
+        
+        # Ultimate fallback - rest
+        return AIDecision(
+            action=ActionType.REST,
+            reason="No valid action found, resting",
+            bdi_info={"mode": "fallback"},
+        )
+    
+    def _convert_to_decision(
+        self,
+        goap_action: "GOAPAction",
+        goal: Optional["Goal"],
+        plan: Optional["Plan"],
+        mode: str = "deliberate",
+    ) -> AIDecision:
+        """Convert a GOAP action to an AIDecision with proper parameters."""
+        action_type = goap_action.action_type
+        target_resource = goap_action.target_resource
+        
+        # Build reason string
+        if goal:
+            reason = f"{goal.name}: {goap_action.name}"
+        else:
+            reason = f"Reactive: {goap_action.name}"
+        
+        # BDI debug info
+        bdi_info = {
+            "mode": mode,
+            "dominant_desire": self.desires.dominant_desire.value if self.desires.dominant_desire else None,
+            "commitment": self.intentions.commitment_strategy.value,
+            "has_intention": self.intentions.has_intention(),
+        }
+        
+        # Handle different action types
+        if action_type == ActionType.CONSUME:
+            return AIDecision(
+                action=action_type,
+                target_resource=target_resource,
+                reason=reason,
+                goal_name=goal.name if goal else "",
+                plan_length=len(plan.actions) if plan else 0,
+                bdi_info=bdi_info,
+            )
+        
+        elif action_type == ActionType.TRADE:
+            return self._create_trade_decision(goap_action, goal, plan, reason, bdi_info)
+        
+        elif action_type in [ActionType.HUNT, ActionType.GATHER, ActionType.CHOP_WOOD,
+                            ActionType.GET_WATER, ActionType.WEAVE]:
+            return AIDecision(
+                action=action_type,
+                target_resource=target_resource,
+                reason=reason,
+                goal_name=goal.name if goal else "",
+                plan_length=len(plan.actions) if plan else 0,
+                bdi_info=bdi_info,
+            )
+        
+        elif action_type == ActionType.BUILD_FIRE:
+            return AIDecision(
+                action=action_type,
+                target_resource=ResourceType.WOOD,
+                reason=reason,
+                goal_name=goal.name if goal else "",
+                plan_length=len(plan.actions) if plan else 0,
+                bdi_info=bdi_info,
+            )
+        
+        elif action_type in [ActionType.REST, ActionType.SLEEP]:
+            return AIDecision(
+                action=action_type,
+                reason=reason,
+                goal_name=goal.name if goal else "",
+                plan_length=len(plan.actions) if plan else 0,
+                bdi_info=bdi_info,
+            )
+        
+        # Default case
+        return AIDecision(
+            action=action_type,
+            target_resource=target_resource,
+            reason=reason,
+            goal_name=goal.name if goal else "",
+            plan_length=len(plan.actions) if plan else 0,
+            bdi_info=bdi_info,
+        )
+    
+    def _create_trade_decision(
+        self,
+        goap_action: "GOAPAction",
+        goal: Optional["Goal"],
+        plan: Optional["Plan"],
+        reason: str,
+        bdi_info: dict,
+    ) -> AIDecision:
+        """Create a trade decision with actual market parameters."""
+        target_resource = goap_action.target_resource
+        action_name = goap_action.name.lower()
+        
+        if "buy" in action_name:
+            # Find the best order to buy from
+            order = self.market.get_cheapest_order(target_resource)
+            
+            if order and order.seller_id != self.agent.id:
+                # Check trust for this seller
+                trust = self.beliefs.get_trade_trust(order.seller_id)
+                
+                # Skip distrusted sellers if we're picky
+                if trust < -0.5 and self.p.price_sensitivity > 1.2:
+                    # Try next cheapest? For now, fall back to gathering
+                    return self._create_gather_fallback(target_resource, reason, goal, plan, bdi_info)
+                
+                # Calculate quantity to buy
+                can_afford = self.agent.money // max(1, order.price_per_unit)
+                space = self.agent.inventory_space()
+                quantity = min(2, can_afford, space, order.quantity)
+                
+                if quantity > 0:
+                    return AIDecision(
+                        action=ActionType.TRADE,
+                        target_resource=target_resource,
+                        order_id=order.id,
+                        quantity=quantity,
+                        price=order.price_per_unit,
+                        reason=f"{reason} @ {order.price_per_unit}c",
+                        goal_name=goal.name if goal else "",
+                        plan_length=len(plan.actions) if plan else 0,
+                        bdi_info=bdi_info,
+                    )
+            
+            # Can't buy - fallback to gathering
+            return self._create_gather_fallback(target_resource, reason, goal, plan, bdi_info)
+        
+        elif "sell" in action_name:
+            # Create a sell order
+            quantity_available = self.agent.get_resource_count(target_resource)
+            
+            # Calculate minimum to keep
+            min_keep = self._get_min_keep(target_resource)
+            quantity_to_sell = min(3, quantity_available - min_keep)
+            
+            if quantity_to_sell > 0:
+                price = self._calculate_sell_price(target_resource)
+                
+                return AIDecision(
+                    action=ActionType.TRADE,
+                    target_resource=target_resource,
+                    quantity=quantity_to_sell,
+                    price=price,
+                    reason=f"{reason} @ {price}c",
+                    goal_name=goal.name if goal else "",
+                    plan_length=len(plan.actions) if plan else 0,
+                    bdi_info=bdi_info,
+                )
+        
+        # Invalid trade action - rest
+        return AIDecision(
+            action=ActionType.REST,
+            reason="Trade not possible",
+            bdi_info=bdi_info,
+        )
+    
+    def _create_gather_fallback(
+        self,
+        resource_type: ResourceType,
+        reason: str,
+        goal: Optional["Goal"],
+        plan: Optional["Plan"],
+        bdi_info: dict,
+    ) -> AIDecision:
+        """Create a gather action as fallback when buying isn't possible."""
+        action_map = {
+            ResourceType.WATER: ActionType.GET_WATER,
+            ResourceType.BERRIES: ActionType.GATHER,
+            ResourceType.MEAT: ActionType.HUNT,
+            ResourceType.WOOD: ActionType.CHOP_WOOD,
+        }
+        
+        action = action_map.get(resource_type, ActionType.GATHER)
+        
+        return AIDecision(
+            action=action,
+            target_resource=resource_type,
+            reason=f"{reason} (gathering instead)",
+            goal_name=goal.name if goal else "",
+            plan_length=len(plan.actions) if plan else 0,
+            bdi_info=bdi_info,
+        )
+    
+    def _get_min_keep(self, resource_type: ResourceType) -> int:
+        """Get minimum quantity to keep for survival."""
+        # Adjusted by hoarding rate from desires
+        hoarding_mult = 0.5 + self.p.hoarding_rate
+        
+        base_min = {
+            ResourceType.WATER: 2,
+            ResourceType.MEAT: 1,
+            ResourceType.BERRIES: 2,
+            ResourceType.WOOD: 1,
+            ResourceType.HIDE: 0,
+        }
+        
+        return int(base_min.get(resource_type, 1) * hoarding_mult)
+    
+    def _calculate_sell_price(self, resource_type: ResourceType) -> int:
+        """Calculate sell price based on fair value and market conditions."""
+        from backend.core.ai import get_energy_cost
+        from backend.config import get_config
+        
+        config = get_config()
+        economy = getattr(config, 'economy', None)
+        energy_to_money_ratio = getattr(economy, 'energy_to_money_ratio', 150) if economy else 150
+        min_price = getattr(economy, 'min_price', 100) if economy else 100
+        
+        energy_cost = get_energy_cost(resource_type)
+        fair_value = max(min_price, int(energy_cost * energy_to_money_ratio))
+        
+        # Apply trading skill
+        sell_modifier = get_trade_price_modifier(self.skills.trading, is_buying=False)
+        
+        # Get market signal
+        signal = self.market.get_market_signal(resource_type)
+        
+        if signal == "sell":  # Scarcity
+            price = int(fair_value * 1.3 * sell_modifier)
+        elif signal == "hold":
+            price = int(fair_value * sell_modifier)
+        else:  # Surplus
+            cheapest = self.market.get_cheapest_order(resource_type)
+            if cheapest and cheapest.seller_id != self.agent.id:
+                price = max(min_price, cheapest.price_per_unit - 1)
+            else:
+                price = int(fair_value * 0.8 * sell_modifier)
+        
+        return max(min_price, price)
+    
+    def record_action_result(self, success: bool, action_type: str) -> None:
+        """Record the result of an action for learning and intention tracking."""
+        # Update intention
+        self.intentions.advance_intention(success)
+        
+        # Update beliefs/memory
+        if success:
+            self.beliefs.record_successful_action(action_type)
+        else:
+            self.beliefs.record_failed_action(action_type)
+
+
+# Persistent BDI state storage for agents
+_agent_bdi_state: dict[str, tuple[BeliefBase, DesireManager, IntentionManager]] = {}
+
+
+def get_bdi_decision(
+    agent: "Agent",
+    market: "OrderBook",
+    step_in_day: int = 1,
+    day_steps: int = 10,
+    current_turn: int = 0,
+    is_night: bool = False,
+) -> AIDecision:
+    """Get a BDI-based AI decision for an agent.
+    
+    This is the main entry point for the BDI AI system.
+    It maintains persistent BDI state for each agent.
+    """
+    # Get or create persistent BDI state
+    if agent.id not in _agent_bdi_state:
+        beliefs = BeliefBase()
+        desires = DesireManager(agent.personality)
+        intentions = IntentionManager.from_personality(agent.personality)
+        _agent_bdi_state[agent.id] = (beliefs, desires, intentions)
+    else:
+        beliefs, desires, intentions = _agent_bdi_state[agent.id]
+    
+    # Create AI instance with persistent state
+    ai = BDIAgentAI(
+        agent=agent,
+        market=market,
+        step_in_day=step_in_day,
+        day_steps=day_steps,
+        current_turn=current_turn,
+        is_night=is_night,
+        beliefs=beliefs,
+        desires=desires,
+        intentions=intentions,
+    )
+    
+    return ai.decide()
+
+
+def reset_bdi_state() -> None:
+    """Reset all BDI state (call on simulation reset)."""
+    global _agent_bdi_state
+    _agent_bdi_state.clear()
+    BDIAgentAI.reset_caches()
+
+
+def remove_agent_bdi_state(agent_id: str) -> None:
+    """Remove BDI state for a specific agent (call on agent death)."""
+    _agent_bdi_state.pop(agent_id, None)

backend/core/bdi/belief.py

@@ -0,0 +1,401 @@
+"""Belief System for BDI agents.
+
+The BeliefBase maintains persistent state about the world from the agent's
+perspective, including:
+- Current sensory data (vitals, inventory, market)
+- Memory of past events (failed trades, good hunting spots)
+- Dirty flags for efficient updates
+
+This replaces the transient WorldState creation with a persistent belief system.
+"""
+
+from dataclasses import dataclass, field
+from typing import TYPE_CHECKING, Optional
+from collections import deque
+
+if TYPE_CHECKING:
+    from backend.domain.agent import Agent
+    from backend.core.market import OrderBook
+
+
+@dataclass
+class MemoryEvent:
+    """A remembered event that may influence future decisions."""
+    event_type: str  # "trade_failed", "hunt_failed", "good_deal", etc.
+    turn: int
+    data: dict = field(default_factory=dict)
+    relevance: float = 1.0  # Decays over time
+
+
+@dataclass
+class BeliefBase:
+    """Persistent belief system for a BDI agent.
+    
+    Maintains both current perceptions and memories of past events.
+    Uses dirty flags to avoid unnecessary recomputation.
+    """
+    
+    # Current perception state (cached WorldState fields)
+    thirst_pct: float = 1.0
+    hunger_pct: float = 1.0
+    heat_pct: float = 1.0
+    energy_pct: float = 1.0
+    
+    # Resource counts
+    water_count: int = 0
+    food_count: int = 0
+    meat_count: int = 0
+    berries_count: int = 0
+    wood_count: int = 0
+    hide_count: int = 0
+    
+    # Inventory state
+    has_clothes: bool = False
+    inventory_space: int = 0
+    inventory_full: bool = False
+    
+    # Economic state
+    money: int = 0
+    is_wealthy: bool = False
+    
+    # Market beliefs (what we believe about market availability)
+    can_buy_water: bool = False
+    can_buy_food: bool = False
+    can_buy_meat: bool = False
+    can_buy_berries: bool = False
+    can_buy_wood: bool = False
+    water_market_price: int = 0
+    food_market_price: int = 0
+    wood_market_price: int = 0
+    
+    # Time beliefs
+    is_night: bool = False
+    is_evening: bool = False
+    step_in_day: int = 0
+    day_steps: int = 10
+    current_turn: int = 0
+    
+    # Personality (cached from agent, rarely changes)
+    wealth_desire: float = 0.5
+    hoarding_rate: float = 0.5
+    risk_tolerance: float = 0.5
+    market_affinity: float = 0.5
+    is_trader: bool = False
+    gather_preference: float = 1.0
+    hunt_preference: float = 1.0
+    trade_preference: float = 1.0
+    
+    # Skills
+    hunting_skill: float = 1.0
+    gathering_skill: float = 1.0
+    trading_skill: float = 1.0
+    
+    # Thresholds (from config)
+    critical_threshold: float = 0.25
+    low_threshold: float = 0.45
+    
+    # Calculated urgencies
+    thirst_urgency: float = 0.0
+    hunger_urgency: float = 0.0
+    heat_urgency: float = 0.0
+    energy_urgency: float = 0.0
+    
+    # === BDI Extensions ===
+    
+    # Memory system - stores past events that influence decisions
+    memories: deque = field(default_factory=lambda: deque(maxlen=50))
+    
+    # Track failed actions for this resource type (turn -> count)
+    failed_hunts: int = 0
+    failed_gathers: int = 0
+    failed_trades: int = 0
+    
+    # Track successful trades with agents (agent_id -> positive count)
+    trusted_traders: dict = field(default_factory=dict)
+    # Track failed trades with agents (agent_id -> negative count)
+    distrusted_traders: dict = field(default_factory=dict)
+    
+    # Dirty flags for optimization
+    _vitals_dirty: bool = True
+    _inventory_dirty: bool = True
+    _market_dirty: bool = True
+    _last_update_turn: int = -1
+    
+    def update_from_sensors(
+        self,
+        agent: "Agent",
+        market: "OrderBook",
+        step_in_day: int = 1,
+        day_steps: int = 10,
+        current_turn: int = 0,
+        is_night: bool = False,
+    ) -> None:
+        """Update beliefs from current agent and market state.
+        
+        Uses dirty flags to skip unchanged data.
+        """
+        from backend.domain.resources import ResourceType
+        from backend.config import get_config
+        
+        self.current_turn = current_turn
+        self.step_in_day = step_in_day
+        self.day_steps = day_steps
+        self.is_night = is_night
+        self.is_evening = step_in_day >= day_steps - 2
+        
+        # Always update vitals (they change every turn)
+        stats = agent.stats
+        self.thirst_pct = stats.thirst / stats.MAX_THIRST
+        self.hunger_pct = stats.hunger / stats.MAX_HUNGER
+        self.heat_pct = stats.heat / stats.MAX_HEAT
+        self.energy_pct = stats.energy / stats.MAX_ENERGY
+        
+        # Update inventory
+        self.water_count = agent.get_resource_count(ResourceType.WATER)
+        self.meat_count = agent.get_resource_count(ResourceType.MEAT)
+        self.berries_count = agent.get_resource_count(ResourceType.BERRIES)
+        self.wood_count = agent.get_resource_count(ResourceType.WOOD)
+        self.hide_count = agent.get_resource_count(ResourceType.HIDE)
+        self.food_count = self.meat_count + self.berries_count
+        
+        self.has_clothes = agent.has_clothes()
+        self.inventory_space = agent.inventory_space()
+        self.inventory_full = agent.inventory_full()
+        self.money = agent.money
+        
+        # Update personality (cached, rarely changes)
+        self.wealth_desire = agent.personality.wealth_desire
+        self.hoarding_rate = agent.personality.hoarding_rate
+        self.risk_tolerance = agent.personality.risk_tolerance
+        self.market_affinity = agent.personality.market_affinity
+        self.gather_preference = agent.personality.gather_preference
+        self.hunt_preference = agent.personality.hunt_preference
+        self.trade_preference = agent.personality.trade_preference
+        
+        # Skills
+        self.hunting_skill = agent.skills.hunting
+        self.gathering_skill = agent.skills.gathering
+        self.trading_skill = agent.skills.trading
+        
+        # Market availability
+        def get_market_info(resource_type: ResourceType) -> tuple[bool, int]:
+            order = market.get_cheapest_order(resource_type)
+            if order and order.seller_id != agent.id and agent.money >= order.price_per_unit:
+                return True, order.price_per_unit
+            return False, 0
+        
+        self.can_buy_water, self.water_market_price = get_market_info(ResourceType.WATER)
+        self.can_buy_meat, meat_price = get_market_info(ResourceType.MEAT)
+        self.can_buy_berries, berries_price = get_market_info(ResourceType.BERRIES)
+        self.can_buy_wood, self.wood_market_price = get_market_info(ResourceType.WOOD)
+        
+        self.can_buy_food = self.can_buy_meat or self.can_buy_berries
+        food_price = min(
+            meat_price if self.can_buy_meat else float('inf'),
+            berries_price if self.can_buy_berries else float('inf')
+        )
+        self.food_market_price = int(food_price) if food_price != float('inf') else 0
+        
+        # Wealth calculation
+        config = get_config()
+        economy_config = getattr(config, 'economy', None)
+        min_wealth_target = getattr(economy_config, 'min_wealth_target', 5000) if economy_config else 5000
+        wealth_target = int(min_wealth_target * (0.5 + self.wealth_desire))
+        self.is_wealthy = self.money >= wealth_target
+        
+        # Trader check
+        self.is_trader = self.trade_preference > 1.3 and self.market_affinity > 0.5
+        
+        # Config thresholds
+        agent_config = config.agent_stats
+        self.critical_threshold = agent_config.critical_threshold
+        self.low_threshold = 0.45
+        
+        # Calculate urgencies
+        self._calculate_urgencies()
+        
+        # Decay old memories
+        self._decay_memories()
+        
+        self._last_update_turn = current_turn
+    
+    def _calculate_urgencies(self) -> None:
+        """Calculate urgency values for each vital stat."""
+        def calc_urgency(pct: float, critical: float, low: float) -> float:
+            if pct >= low:
+                return 0.0
+            elif pct >= critical:
+                return 1.0 - (pct - critical) / (low - critical)
+            else:
+                return 1.0 + (critical - pct) / critical * 2.0
+        
+        self.thirst_urgency = calc_urgency(self.thirst_pct, self.critical_threshold, self.low_threshold)
+        self.hunger_urgency = calc_urgency(self.hunger_pct, self.critical_threshold, self.low_threshold)
+        self.heat_urgency = calc_urgency(self.heat_pct, self.critical_threshold, self.low_threshold)
+        
+        if self.energy_pct < 0.25:
+            self.energy_urgency = 2.0
+        elif self.energy_pct < 0.40:
+            self.energy_urgency = 1.0
+        else:
+            self.energy_urgency = 0.0
+    
+    def _decay_memories(self) -> None:
+        """Decay memory relevance over time."""
+        for memory in self.memories:
+            age = self.current_turn - memory.turn
+            # Memories decay exponentially with age
+            memory.relevance = max(0.1, 1.0 / (1.0 + age * 0.1))
+    
+    def add_memory(self, event_type: str, data: dict = None) -> None:
+        """Add a new memory event."""
+        self.memories.append(MemoryEvent(
+            event_type=event_type,
+            turn=self.current_turn,
+            data=data or {},
+            relevance=1.0,
+        ))
+    
+    def record_failed_action(self, action_type: str) -> None:
+        """Record a failed action for learning."""
+        if action_type == "hunt":
+            self.failed_hunts += 1
+            self.add_memory("hunt_failed")
+        elif action_type == "gather":
+            self.failed_gathers += 1
+            self.add_memory("gather_failed")
+        elif action_type == "trade":
+            self.failed_trades += 1
+            self.add_memory("trade_failed")
+    
+    def record_successful_action(self, action_type: str) -> None:
+        """Record a successful action, reducing failure counts."""
+        if action_type == "hunt":
+            self.failed_hunts = max(0, self.failed_hunts - 1)
+        elif action_type == "gather":
+            self.failed_gathers = max(0, self.failed_gathers - 1)
+        elif action_type == "trade":
+            self.failed_trades = max(0, self.failed_trades - 1)
+    
+    def record_trade_partner(self, partner_id: str, success: bool) -> None:
+        """Track trade relationship with another agent."""
+        if success:
+            self.trusted_traders[partner_id] = self.trusted_traders.get(partner_id, 0) + 1
+            # Reduce distrust if they've been bad before
+            if partner_id in self.distrusted_traders:
+                self.distrusted_traders[partner_id] = max(0, self.distrusted_traders[partner_id] - 1)
+        else:
+            self.distrusted_traders[partner_id] = self.distrusted_traders.get(partner_id, 0) + 1
+    
+    def get_trade_trust(self, partner_id: str) -> float:
+        """Get trust level for a trade partner (-1 to 1)."""
+        trust = self.trusted_traders.get(partner_id, 0)
+        distrust = self.distrusted_traders.get(partner_id, 0)
+        total = trust + distrust
+        if total == 0:
+            return 0.0  # Unknown partner
+        return (trust - distrust) / total
+    
+    def has_critical_need(self) -> bool:
+        """Check if any vital stat is critical (requires immediate attention)."""
+        return (
+            self.thirst_urgency >= 2.0 or
+            self.hunger_urgency >= 2.0 or
+            self.heat_urgency >= 2.0 or
+            self.energy_urgency >= 2.0
+        )
+    
+    def get_most_urgent_need(self) -> Optional[str]:
+        """Get the most urgent vital need, if any."""
+        urgencies = {
+            "thirst": self.thirst_urgency,
+            "hunger": self.hunger_urgency,
+            "heat": self.heat_urgency,
+            "energy": self.energy_urgency,
+        }
+        max_urgency = max(urgencies.values())
+        if max_urgency < 0.5:
+            return None
+        return max(urgencies, key=urgencies.get)
+    
+    def to_world_state(self):
+        """Convert beliefs to a WorldState for GOAP planner compatibility."""
+        from backend.core.goap.world_state import WorldState
+
+        return WorldState(
+            thirst_pct=self.thirst_pct,
+            hunger_pct=self.hunger_pct,
+            heat_pct=self.heat_pct,
+            energy_pct=self.energy_pct,
+            water_count=self.water_count,
+            food_count=self.food_count,
+            meat_count=self.meat_count,
+            berries_count=self.berries_count,
+            wood_count=self.wood_count,
+            hide_count=self.hide_count,
+            has_clothes=self.has_clothes,
+            inventory_space=self.inventory_space,
+            inventory_full=self.inventory_full,
+            money=self.money,
+            is_wealthy=self.is_wealthy,
+            can_buy_water=self.can_buy_water,
+            can_buy_food=self.can_buy_food,
+            can_buy_meat=self.can_buy_meat,
+            can_buy_berries=self.can_buy_berries,
+            can_buy_wood=self.can_buy_wood,
+            water_market_price=self.water_market_price,
+            food_market_price=self.food_market_price,
+            wood_market_price=self.wood_market_price,
+            is_night=self.is_night,
+            is_evening=self.is_evening,
+            step_in_day=self.step_in_day,
+            day_steps=self.day_steps,
+            wealth_desire=self.wealth_desire,
+            hoarding_rate=self.hoarding_rate,
+            risk_tolerance=self.risk_tolerance,
+            market_affinity=self.market_affinity,
+            is_trader=self.is_trader,
+            gather_preference=self.gather_preference,
+            hunt_preference=self.hunt_preference,
+            trade_preference=self.trade_preference,
+            hunting_skill=self.hunting_skill,
+            gathering_skill=self.gathering_skill,
+            trading_skill=self.trading_skill,
+            critical_threshold=self.critical_threshold,
+            low_threshold=self.low_threshold,
+        )
+    
+    def to_dict(self) -> dict:
+        """Convert to dictionary for debugging/logging."""
+        return {
+            "vitals": {
+                "thirst": round(self.thirst_pct, 2),
+                "hunger": round(self.hunger_pct, 2),
+                "heat": round(self.heat_pct, 2),
+                "energy": round(self.energy_pct, 2),
+            },
+            "urgencies": {
+                "thirst": round(self.thirst_urgency, 2),
+                "hunger": round(self.hunger_urgency, 2),
+                "heat": round(self.heat_urgency, 2),
+                "energy": round(self.energy_urgency, 2),
+            },
+            "inventory": {
+                "water": self.water_count,
+                "meat": self.meat_count,
+                "berries": self.berries_count,
+                "wood": self.wood_count,
+                "hide": self.hide_count,
+                "space": self.inventory_space,
+            },
+            "economy": {
+                "money": self.money,
+                "is_wealthy": self.is_wealthy,
+            },
+            "memory": {
+                "failed_hunts": self.failed_hunts,
+                "failed_gathers": self.failed_gathers,
+                "failed_trades": self.failed_trades,
+                "memory_count": len(self.memories),
+            },
+        }

backend/core/bdi/desire.py

@@ -0,0 +1,347 @@
+"""Desire System for BDI agents.
+
+Desires represent high-level, long-term motivations that persist across turns.
+Unlike GOAP goals (which are immediate targets), desires are personality-driven
+and influence which goals get activated.
+
+Key concepts:
+- Desires are weighted by personality traits
+- Desires can be satisfied temporarily but return
+- Desires influence goal selection, not direct actions
+"""
+
+from dataclasses import dataclass
+from enum import Enum
+from typing import TYPE_CHECKING, Optional
+
+if TYPE_CHECKING:
+    from backend.core.bdi.belief import BeliefBase
+    from backend.domain.personality import PersonalityTraits
+    from backend.core.goap.goal import Goal
+
+
+class DesireType(Enum):
+    """Types of high-level desires."""
+    SURVIVAL = "survival"           # Stay alive (thirst, hunger, heat, energy)
+    ACCUMULATE_WEALTH = "wealth"    # Build up money reserves
+    STOCK_RESOURCES = "stock"       # Hoard resources for security
+    MASTER_PROFESSION = "mastery"   # Improve skills in preferred activity
+    SOCIAL_STANDING = "social"      # Gain reputation through trade
+    COMFORT = "comfort"             # Maintain clothes, warmth, rest
+
+
+@dataclass
+class Desire:
+    """A high-level motivation that influences goal selection.
+    
+    Desires have:
+    - A base intensity (how strongly the agent wants this)
+    - A satisfaction level (0-1, how fulfilled is this desire currently)
+    - A personality weight (how much this agent's personality cares)
+    """
+    desire_type: DesireType
+    base_intensity: float = 1.0      # Base importance (0-2)
+    satisfaction: float = 0.5        # Current fulfillment (0-1)
+    personality_weight: float = 1.0  # Multiplier from personality
+    
+    # Persistence tracking
+    turns_pursued: int = 0           # How long we've been chasing this
+    turns_since_progress: int = 0    # Turns since we made progress
+    
+    @property
+    def effective_intensity(self) -> float:
+        """Calculate current desire intensity considering satisfaction."""
+        # Desire is stronger when unsatisfied
+        unsatisfied = 1.0 - self.satisfaction
+        # Apply personality weight
+        intensity = self.base_intensity * self.personality_weight * (0.5 + unsatisfied)
+        # Reduce intensity if pursued too long without progress (boredom)
+        if self.turns_since_progress > 10:
+            boredom_factor = max(0.3, 1.0 - (self.turns_since_progress - 10) * 0.05)
+            intensity *= boredom_factor
+        return intensity
+    
+    def update_satisfaction(self, new_satisfaction: float) -> None:
+        """Update satisfaction level and track progress."""
+        if new_satisfaction > self.satisfaction:
+            self.turns_since_progress = 0  # Made progress!
+        else:
+            self.turns_since_progress += 1
+        self.satisfaction = max(0.0, min(1.0, new_satisfaction))
+        self.turns_pursued += 1
+    
+    def reset_pursuit(self) -> None:
+        """Reset pursuit tracking (when switching to different desire)."""
+        self.turns_pursued = 0
+        self.turns_since_progress = 0
+
+
+class DesireManager:
+    """Manages an agent's desires and converts them to active goals.
+    
+    The DesireManager:
+    1. Maintains a set of desires weighted by personality
+    2. Updates desire satisfaction based on beliefs
+    3. Selects which desires should drive current goals
+    4. Filters/prioritizes GOAP goals based on active desires
+    """
+    
+    def __init__(self, personality: "PersonalityTraits"):
+        """Initialize desires based on personality."""
+        self.desires: dict[DesireType, Desire] = {}
+        self._init_desires(personality)
+        
+        # Track which desire is currently dominant
+        self.dominant_desire: Optional[DesireType] = None
+        self.stubbornness: float = 0.5  # How reluctant to switch desires
+    
+    def _init_desires(self, personality: "PersonalityTraits") -> None:
+        """Initialize desires with personality-based weights."""
+        # Survival - everyone has this, but risk-averse agents weight it higher
+        self.desires[DesireType.SURVIVAL] = Desire(
+            desire_type=DesireType.SURVIVAL,
+            base_intensity=2.0,  # Always high base
+            personality_weight=1.5 - personality.risk_tolerance * 0.5,
+        )
+        
+        # Accumulate Wealth - driven by wealth_desire
+        self.desires[DesireType.ACCUMULATE_WEALTH] = Desire(
+            desire_type=DesireType.ACCUMULATE_WEALTH,
+            base_intensity=1.0,
+            personality_weight=0.5 + personality.wealth_desire,
+        )
+        
+        # Stock Resources - driven by hoarding_rate
+        self.desires[DesireType.STOCK_RESOURCES] = Desire(
+            desire_type=DesireType.STOCK_RESOURCES,
+            base_intensity=0.8,
+            personality_weight=0.5 + personality.hoarding_rate,
+        )
+        
+        # Master Profession - driven by strongest preference
+        max_pref = max(
+            personality.hunt_preference,
+            personality.gather_preference,
+            personality.trade_preference,
+        )
+        self.desires[DesireType.MASTER_PROFESSION] = Desire(
+            desire_type=DesireType.MASTER_PROFESSION,
+            base_intensity=0.6,
+            personality_weight=max_pref * 0.5,
+        )
+        
+        # Social Standing - driven by market_affinity and trade_preference
+        self.desires[DesireType.SOCIAL_STANDING] = Desire(
+            desire_type=DesireType.SOCIAL_STANDING,
+            base_intensity=0.5,
+            personality_weight=personality.market_affinity * personality.trade_preference * 0.5,
+        )
+        
+        # Comfort - moderate for everyone
+        self.desires[DesireType.COMFORT] = Desire(
+            desire_type=DesireType.COMFORT,
+            base_intensity=0.7,
+            personality_weight=1.0 - personality.risk_tolerance * 0.3,
+        )
+        
+        # Calculate stubbornness from personality
+        # High hoarding + low risk tolerance = stubborn
+        self.stubbornness = (personality.hoarding_rate + (1.0 - personality.risk_tolerance)) / 2
+    
+    def update_from_beliefs(self, beliefs: "BeliefBase") -> None:
+        """Update desire satisfaction based on current beliefs."""
+        # Survival satisfaction based on vitals
+        min_vital = min(
+            beliefs.thirst_pct,
+            beliefs.hunger_pct,
+            beliefs.heat_pct,
+            beliefs.energy_pct,
+        )
+        self.desires[DesireType.SURVIVAL].update_satisfaction(min_vital)
+        
+        # Wealth satisfaction based on money relative to wealthy threshold
+        if beliefs.is_wealthy:
+            self.desires[DesireType.ACCUMULATE_WEALTH].update_satisfaction(0.8)
+        else:
+            # Scale satisfaction by money (rough approximation)
+            wealth_sat = min(1.0, beliefs.money / 10000)
+            self.desires[DesireType.ACCUMULATE_WEALTH].update_satisfaction(wealth_sat)
+        
+        # Stock satisfaction based on inventory
+        total_resources = (
+            beliefs.water_count +
+            beliefs.food_count +
+            beliefs.wood_count
+        )
+        stock_sat = min(1.0, total_resources / 15)  # 15 resources = satisfied
+        self.desires[DesireType.STOCK_RESOURCES].update_satisfaction(stock_sat)
+        
+        # Mastery satisfaction based on skill levels
+        max_skill = max(
+            beliefs.hunting_skill,
+            beliefs.gathering_skill,
+            beliefs.trading_skill,
+        )
+        mastery_sat = (max_skill - 1.0) / 1.0  # 0 at skill 1.0, 1 at skill 2.0
+        self.desires[DesireType.MASTER_PROFESSION].update_satisfaction(max(0, mastery_sat))
+        
+        # Social satisfaction - harder to measure, use trade success
+        trade_success = max(0, 5 - beliefs.failed_trades) / 5
+        self.desires[DesireType.SOCIAL_STANDING].update_satisfaction(trade_success)
+        
+        # Comfort satisfaction
+        comfort_factors = [
+            beliefs.energy_pct,
+            beliefs.heat_pct,
+            1.0 if beliefs.has_clothes else 0.5,
+        ]
+        comfort_sat = sum(comfort_factors) / len(comfort_factors)
+        self.desires[DesireType.COMFORT].update_satisfaction(comfort_sat)
+    
+    def get_active_desires(self) -> list[Desire]:
+        """Get desires sorted by effective intensity (highest first)."""
+        return sorted(
+            self.desires.values(),
+            key=lambda d: d.effective_intensity,
+            reverse=True,
+        )
+    
+    def should_switch_desire(self, new_dominant: DesireType) -> bool:
+        """Determine if we should switch dominant desire.
+        
+        Stubborn agents require a larger intensity difference to switch.
+        """
+        if self.dominant_desire is None:
+            return True
+        
+        if new_dominant == self.dominant_desire:
+            return False
+        
+        current = self.desires[self.dominant_desire]
+        new = self.desires[new_dominant]
+        
+        # Calculate required intensity difference based on stubbornness
+        # Stubbornness 0.5 = need 20% more intensity to switch
+        # Stubbornness 1.0 = need 50% more intensity to switch
+        required_difference = 1.0 + self.stubbornness * 0.5
+        
+        return new.effective_intensity > current.effective_intensity * required_difference
+    
+    def update_dominant_desire(self) -> DesireType:
+        """Update and return the currently dominant desire."""
+        active = self.get_active_desires()
+        if not active:
+            return DesireType.SURVIVAL
+        
+        top_desire = active[0].desire_type
+        
+        if self.should_switch_desire(top_desire):
+            # Reset pursuit on old desire if switching
+            if self.dominant_desire and self.dominant_desire != top_desire:
+                self.desires[self.dominant_desire].reset_pursuit()
+            self.dominant_desire = top_desire
+        
+        return self.dominant_desire
+    
+    def filter_goals_by_desire(
+        self,
+        goals: list["Goal"],
+        beliefs: "BeliefBase",
+    ) -> list["Goal"]:
+        """Filter and re-prioritize goals based on active desires.
+        
+        This is where desires influence GOAP goal selection:
+        - Goals aligned with dominant desire get priority boost
+        - Goals conflicting with desires get reduced priority
+        """
+        # Update dominant desire
+        dominant = self.update_dominant_desire()
+        
+        # Get world state for goal priority calculation
+        world_state = beliefs.to_world_state()
+        
+        # Calculate goal priorities with desire modifiers
+        goal_priorities: list[tuple["Goal", float]] = []
+        
+        for goal in goals:
+            base_priority = goal.get_priority(world_state)
+            
+            # Apply desire-based modifiers
+            modifier = self._get_goal_desire_modifier(goal.name, dominant, beliefs)
+            adjusted_priority = base_priority * modifier
+            
+            goal_priorities.append((goal, adjusted_priority))
+        
+        # Sort by adjusted priority
+        goal_priorities.sort(key=lambda x: x[1], reverse=True)
+        
+        # Return goals (the priorities will be recalculated by planner,
+        # but we've filtered/ordered them by our preferences)
+        return [g for g, _ in goal_priorities]
+    
+    def _get_goal_desire_modifier(
+        self,
+        goal_name: str,
+        dominant_desire: DesireType,
+        beliefs: "BeliefBase",
+    ) -> float:
+        """Get priority modifier for a goal based on dominant desire."""
+        goal_lower = goal_name.lower()
+        
+        # Map goals to desires they serve
+        goal_desire_map = {
+            # Survival goals
+            "satisfy thirst": DesireType.SURVIVAL,
+            "satisfy hunger": DesireType.SURVIVAL,
+            "maintain heat": DesireType.SURVIVAL,
+            "restore energy": DesireType.SURVIVAL,
+            # Wealth goals
+            "build wealth": DesireType.ACCUMULATE_WEALTH,
+            "sell excess": DesireType.ACCUMULATE_WEALTH,
+            "find deals": DesireType.ACCUMULATE_WEALTH,
+            "trader arbitrage": DesireType.ACCUMULATE_WEALTH,
+            # Stock goals
+            "stock water": DesireType.STOCK_RESOURCES,
+            "stock food": DesireType.STOCK_RESOURCES,
+            "stock wood": DesireType.STOCK_RESOURCES,
+            # Comfort goals
+            "get clothes": DesireType.COMFORT,
+            "sleep": DesireType.COMFORT,
+        }
+        
+        # Find which desire this goal serves
+        goal_desire = None
+        for gn, desire in goal_desire_map.items():
+            if gn in goal_lower:
+                goal_desire = desire
+                break
+        
+        if goal_desire is None:
+            return 1.0  # No modifier for unknown goals
+        
+        # Boost if aligned with dominant desire
+        if goal_desire == dominant_desire:
+            return 1.3 + self.desires[dominant_desire].effective_intensity * 0.2
+        
+        # Survival always gets a baseline boost if critical
+        if goal_desire == DesireType.SURVIVAL and beliefs.has_critical_need():
+            return 2.0  # Critical needs override other desires
+        
+        # Slight reduction for non-aligned goals
+        return 0.9
+    
+    def to_dict(self) -> dict:
+        """Convert to dictionary for debugging/logging."""
+        return {
+            "dominant_desire": self.dominant_desire.value if self.dominant_desire else None,
+            "stubbornness": round(self.stubbornness, 2),
+            "desires": {
+                d.desire_type.value: {
+                    "intensity": round(d.effective_intensity, 2),
+                    "satisfaction": round(d.satisfaction, 2),
+                    "personality_weight": round(d.personality_weight, 2),
+                    "turns_pursued": d.turns_pursued,
+                }
+                for d in self.desires.values()
+            },
+        }

backend/core/bdi/intention.py

@@ -0,0 +1,289 @@
+"""Intention System for BDI agents.
+
+Intentions represent committed plans that the agent is executing.
+Unlike desires (motivations) or goals (targets), intentions are
+concrete action sequences the agent has decided to pursue.
+
+Key concepts:
+- Intention persistence: agents stick to plans unless interrupted
+- Commitment strategies: different levels of plan commitment
+- Plan monitoring: detecting when a plan becomes invalid
+"""
+
+from dataclasses import dataclass
+from enum import Enum
+from typing import TYPE_CHECKING, Optional
+
+if TYPE_CHECKING:
+    from backend.core.bdi.belief import BeliefBase
+    from backend.core.bdi.desire import DesireManager
+    from backend.core.goap.goal import Goal
+    from backend.core.goap.action import GOAPAction
+    from backend.core.goap.planner import Plan
+
+
+class CommitmentStrategy(Enum):
+    """How strongly an agent commits to their current intention."""
+    REACTIVE = "reactive"       # Replan every turn (no commitment)
+    CAUTIOUS = "cautious"       # Replan if priorities shift significantly
+    DETERMINED = "determined"   # Stick to plan unless it becomes impossible
+    STUBBORN = "stubborn"       # Only abandon for critical interrupts
+
+
+@dataclass
+class Intention:
+    """A committed plan that the agent is executing.
+    
+    Tracks the goal, plan, and execution progress.
+    """
+    goal: "Goal"                           # The goal we're pursuing
+    plan: "Plan"                           # The plan to achieve it
+    start_turn: int                        # When we committed
+    actions_completed: int = 0             # How many actions done
+    last_action_success: bool = True       # Did last action succeed?
+    consecutive_failures: int = 0          # Failures in a row
+    
+    @property
+    def current_action(self) -> Optional["GOAPAction"]:
+        """Get the next action to execute."""
+        if self.plan.is_empty:
+            return None
+        remaining = self.plan.actions[self.actions_completed:]
+        return remaining[0] if remaining else None
+    
+    @property
+    def is_complete(self) -> bool:
+        """Check if the plan has been fully executed."""
+        return self.actions_completed >= len(self.plan.actions)
+    
+    @property
+    def remaining_actions(self) -> int:
+        """Number of actions left in the plan."""
+        return max(0, len(self.plan.actions) - self.actions_completed)
+    
+    def advance(self, success: bool) -> None:
+        """Mark current action as executed and advance."""
+        self.last_action_success = success
+        if success:
+            self.actions_completed += 1
+            self.consecutive_failures = 0
+        else:
+            self.consecutive_failures += 1
+
+
+class IntentionManager:
+    """Manages the agent's current intention and commitment.
+    
+    Responsibilities:
+    - Maintain the current intention (goal + plan)
+    - Decide when to continue vs. replan
+    - Handle plan failure and recovery
+    """
+    
+    def __init__(self, commitment_strategy: CommitmentStrategy = CommitmentStrategy.CAUTIOUS):
+        self.current_intention: Optional[Intention] = None
+        self.commitment_strategy = commitment_strategy
+        
+        # Tracking
+        self.intentions_completed: int = 0
+        self.intentions_abandoned: int = 0
+        self.total_actions_executed: int = 0
+        
+        # Thresholds for reconsideration
+        self.max_consecutive_failures: int = 2
+        self.priority_switch_threshold: float = 1.5  # New goal must be 1.5x priority
+    
+    @classmethod
+    def from_personality(cls, personality) -> "IntentionManager":
+        """Create an IntentionManager with commitment strategy based on personality."""
+        # Derive commitment from personality traits
+        # High hoarding + low risk tolerance = stubborn
+        commitment_score = (
+            personality.hoarding_rate * 0.4 +
+            (1.0 - personality.risk_tolerance) * 0.4 +
+            (1.0 - personality.market_affinity) * 0.2
+        )
+        
+        if commitment_score > 0.7:
+            strategy = CommitmentStrategy.STUBBORN
+        elif commitment_score > 0.5:
+            strategy = CommitmentStrategy.DETERMINED
+        elif commitment_score > 0.3:
+            strategy = CommitmentStrategy.CAUTIOUS
+        else:
+            strategy = CommitmentStrategy.REACTIVE
+        
+        return cls(commitment_strategy=strategy)
+    
+    def has_intention(self) -> bool:
+        """Check if we have an active intention."""
+        return (
+            self.current_intention is not None and
+            not self.current_intention.is_complete
+        )
+    
+    def get_next_action(self) -> Optional["GOAPAction"]:
+        """Get the next action from current intention."""
+        if not self.has_intention():
+            return None
+        return self.current_intention.current_action
+    
+    def should_reconsider(
+        self,
+        beliefs: "BeliefBase",
+        desire_manager: "DesireManager",
+        available_goals: list["Goal"],
+    ) -> bool:
+        """Determine if we should reconsider our current intention.
+        
+        This implements the commitment strategy logic.
+        """
+        # No intention = definitely need to plan
+        if not self.has_intention():
+            return True
+        
+        intention = self.current_intention
+        
+        # Check for critical interrupts (always reconsider)
+        if beliefs.has_critical_need():
+            # But only if current intention isn't already addressing it
+            urgent_need = beliefs.get_most_urgent_need()
+            if not self._intention_addresses_need(intention, urgent_need):
+                return True
+        
+        # Check for too many failures
+        if intention.consecutive_failures >= self.max_consecutive_failures:
+            return True
+        
+        # Strategy-specific logic
+        if self.commitment_strategy == CommitmentStrategy.REACTIVE:
+            return True  # Always replan
+        
+        elif self.commitment_strategy == CommitmentStrategy.CAUTIOUS:
+            # Reconsider if a significantly better goal is available
+            return self._better_goal_available(
+                beliefs, desire_manager, available_goals,
+                threshold=self.priority_switch_threshold
+            )
+        
+        elif self.commitment_strategy == CommitmentStrategy.DETERMINED:
+            # Only reconsider for much better goals or if plan is failing
+            return (
+                intention.consecutive_failures > 0 or
+                self._better_goal_available(
+                    beliefs, desire_manager, available_goals,
+                    threshold=2.0  # Need 2x priority to switch
+                )
+            )
+        
+        elif self.commitment_strategy == CommitmentStrategy.STUBBORN:
+            # Only abandon for critical needs or impossible plans
+            return (
+                beliefs.has_critical_need() or
+                intention.consecutive_failures >= self.max_consecutive_failures
+            )
+        
+        return False
+    
+    def _intention_addresses_need(self, intention: Intention, need: str) -> bool:
+        """Check if current intention addresses a vital need."""
+        if not intention or not need:
+            return False
+        
+        goal_name = intention.goal.name.lower()
+        need_map = {
+            "thirst": "thirst",
+            "hunger": "hunger",
+            "heat": "heat",
+            "energy": "energy",
+        }
+        
+        return need_map.get(need, "") in goal_name
+    
+    def _better_goal_available(
+        self,
+        beliefs: "BeliefBase",
+        desire_manager: "DesireManager",
+        available_goals: list["Goal"],
+        threshold: float,
+    ) -> bool:
+        """Check if there's a significantly better goal available."""
+        if not self.has_intention():
+            return True
+        
+        current_goal = self.current_intention.goal
+        world_state = beliefs.to_world_state()
+        current_priority = current_goal.get_priority(world_state)
+        
+        # Get desire-filtered goals
+        filtered_goals = desire_manager.filter_goals_by_desire(available_goals, beliefs)
+        
+        for goal in filtered_goals:
+            if goal.name == current_goal.name:
+                continue
+            
+            goal_priority = goal.get_priority(world_state)
+            if goal_priority > current_priority * threshold:
+                return True
+        
+        return False
+    
+    def commit_to_plan(self, goal: "Goal", plan: "Plan", current_turn: int) -> None:
+        """Commit to a new intention."""
+        # Track abandoned intention
+        if self.has_intention():
+            self.intentions_abandoned += 1
+        
+        self.current_intention = Intention(
+            goal=goal,
+            plan=plan,
+            start_turn=current_turn,
+        )
+    
+    def advance_intention(self, success: bool) -> None:
+        """Record action execution and advance the intention."""
+        if not self.has_intention():
+            return
+        
+        self.current_intention.advance(success)
+        self.total_actions_executed += 1
+        
+        # Check if intention is complete
+        if self.current_intention.is_complete:
+            self.intentions_completed += 1
+            self.current_intention = None
+    
+    def abandon_intention(self, reason: str = "unknown") -> None:
+        """Abandon the current intention."""
+        if self.has_intention():
+            self.intentions_abandoned += 1
+            self.current_intention = None
+    
+    def get_plan_progress(self) -> dict:
+        """Get current plan execution progress."""
+        if not self.has_intention():
+            return {"has_intention": False}
+        
+        intention = self.current_intention
+        return {
+            "has_intention": True,
+            "goal": intention.goal.name,
+            "actions_completed": intention.actions_completed,
+            "remaining_actions": intention.remaining_actions,
+            "consecutive_failures": intention.consecutive_failures,
+            "turns_active": 0,  # Would need current_turn to calculate
+        }
+    
+    def to_dict(self) -> dict:
+        """Convert to dictionary for debugging/logging."""
+        return {
+            "commitment_strategy": self.commitment_strategy.value,
+            "has_intention": self.has_intention(),
+            "current_goal": self.current_intention.goal.name if self.has_intention() else None,
+            "plan_progress": self.get_plan_progress(),
+            "stats": {
+                "intentions_completed": self.intentions_completed,
+                "intentions_abandoned": self.intentions_abandoned,
+                "total_actions": self.total_actions_executed,
+            },
+        }

backend/core/logger.py

@@ -1,10 +1,17 @@
 """Simulation logger for detailed step-by-step logging.
 
-Performance-optimized: logging can be disabled or reduced via config.
+Performance-optimized with non-blocking I/O:
+- Logging can be disabled or reduced via config
+- File writes happen in a background thread (producer-consumer pattern)
+- Agent lookups use O(1) dict instead of O(n) list search
+- No in-memory accumulation of all entries
 """
 
 import json
 import logging
+import threading
+import queue
+import atexit
 from dataclasses import dataclass, field, asdict
 from datetime import datetime
 from pathlib import Path
@@ -60,12 +67,150 @@ class TurnLogEntry:
         }
 
 
+class AsyncLogWriter:
+    """Background thread that handles file I/O asynchronously.
+    
+    Uses a producer-consumer pattern to decouple log generation
+    from file writing, preventing I/O from blocking the simulation.
+    """
+    
+    def __init__(self, max_queue_size: int = 1000):
+        self._queue: queue.Queue = queue.Queue(maxsize=max_queue_size)
+        self._stop_event = threading.Event()
+        self._thread: Optional[threading.Thread] = None
+        self._files: dict[str, TextIO] = {}
+        self._lock = threading.Lock()
+    
+    def start(self) -> None:
+        """Start the background writer thread."""
+        if self._thread is not None and self._thread.is_alive():
+            return
+        
+        self._stop_event.clear()
+        self._thread = threading.Thread(target=self._writer_loop, daemon=True)
+        self._thread.start()
+    
+    def stop(self, timeout: float = 2.0) -> None:
+        """Stop the background writer thread and flush remaining items."""
+        self._stop_event.set()
+        if self._thread is not None:
+            self._thread.join(timeout=timeout)
+            self._thread = None
+        
+        # Process any remaining items in the queue
+        self._drain_queue()
+        
+        # Close all files
+        with self._lock:
+            for f in self._files.values():
+                try:
+                    f.close()
+                except Exception:
+                    pass
+            self._files.clear()
+    
+    def _writer_loop(self) -> None:
+        """Main loop for the background writer thread."""
+        while not self._stop_event.is_set():
+            try:
+                # Wait for items with timeout to allow stop checks
+                item = self._queue.get(timeout=0.1)
+                self._process_item(item)
+                self._queue.task_done()
+            except queue.Empty:
+                continue
+            except Exception as e:
+                # Log errors but don't crash the thread
+                logging.getLogger("simulation").warning(f"Async log writer error: {e}")
+    
+    def _process_item(self, item: dict) -> None:
+        """Process a single log item."""
+        action = item.get("action")
+        
+        if action == "open":
+            self._open_file(item["path"], item["file_id"])
+        elif action == "write":
+            self._write_to_file(item["file_id"], item["data"])
+        elif action == "flush":
+            self._flush_file(item.get("file_id"))
+        elif action == "close":
+            self._close_file(item["file_id"])
+    
+    def _open_file(self, path: str, file_id: str) -> None:
+        """Open a file for writing."""
+        with self._lock:
+            if file_id not in self._files:
+                self._files[file_id] = open(path, "w", encoding="utf-8")
+    
+    def _write_to_file(self, file_id: str, data: str) -> None:
+        """Write data to a file."""
+        with self._lock:
+            f = self._files.get(file_id)
+            if f:
+                f.write(data)
+    
+    def _flush_file(self, file_id: Optional[str] = None) -> None:
+        """Flush file(s) to disk."""
+        with self._lock:
+            if file_id:
+                f = self._files.get(file_id)
+                if f:
+                    f.flush()
+            else:
+                for f in self._files.values():
+                    f.flush()
+    
+    def _close_file(self, file_id: str) -> None:
+        """Close a file."""
+        with self._lock:
+            f = self._files.pop(file_id, None)
+            if f:
+                f.close()
+    
+    def _drain_queue(self) -> None:
+        """Process all remaining items in the queue."""
+        while True:
+            try:
+                item = self._queue.get_nowait()
+                self._process_item(item)
+                self._queue.task_done()
+            except queue.Empty:
+                break
+    
+    def enqueue(self, item: dict) -> bool:
+        """Add an item to the write queue.
+        
+        Returns False if queue is full (item dropped).
+        """
+        try:
+            self._queue.put_nowait(item)
+            return True
+        except queue.Full:
+            return False
+    
+    def open_file(self, path: str, file_id: str) -> None:
+        """Queue a file open operation."""
+        self.enqueue({"action": "open", "path": path, "file_id": file_id})
+    
+    def write(self, file_id: str, data: str) -> None:
+        """Queue a write operation."""
+        self.enqueue({"action": "write", "file_id": file_id, "data": data})
+    
+    def flush(self, file_id: Optional[str] = None) -> None:
+        """Queue a flush operation."""
+        self.enqueue({"action": "flush", "file_id": file_id})
+    
+    def close_file(self, file_id: str) -> None:
+        """Queue a file close operation."""
+        self.enqueue({"action": "close", "file_id": file_id})
+
+
 class SimulationLogger:
     """Logger that dumps detailed simulation data to files.
 
     Performance optimized:
     - Logging can be disabled entirely via config
-    - File flushing is batched (not every turn)
+    - File writes happen in background thread (non-blocking)
     - Agent lookups use O(1) dict instead of O(n) list search
     - No in-memory accumulation of all entries
     """
@@ -80,7 +225,17 @@ class SimulationLogger:
         self.detailed_logging = perf_config.detailed_logging
         self.flush_interval = perf_config.log_flush_interval
         
-        # File handles (only created if logging enabled)
+        # Get async logging config (default to True if available)
+        self.async_logging = getattr(perf_config, 'async_logging', True)
+
+        # Async writer (only created if logging enabled)
+        self._async_writer: Optional[AsyncLogWriter] = None
+        
+        # File IDs for async writing
+        self._json_file_id: Optional[str] = None
+        self._summary_file_id: Optional[str] = None
+        
+        # Fallback: synchronous file handles (used if async disabled)
         self._json_file: Optional[TextIO] = None
         self._summary_file: Optional[TextIO] = None
 
@@ -96,6 +251,10 @@ class SimulationLogger:
         # Turn counter for flush batching
         self._turns_since_flush = 0
         
+        # Stats
+        self._items_queued = 0
+        self._items_dropped = 0
+
     def start_session(self, config: dict) -> None:
         """Start a new logging session."""
         if not self.logging_enabled:
@@ -103,19 +262,39 @@ class SimulationLogger:
 
         self.log_dir.mkdir(exist_ok=True)
 
-        # Create session-specific log file
+        # Create session-specific log file paths
         timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-        session_file = self.log_dir / f"sim_{timestamp}.jsonl"
+        json_path = self.log_dir / f"sim_{timestamp}.jsonl"
-        summary_file = self.log_dir / f"sim_{timestamp}_summary.txt"
+        summary_path = self.log_dir / f"sim_{timestamp}_summary.txt"
         
-        self._json_file = open(session_file, "w")
+        if self.async_logging:
-        self._summary_file = open(summary_file, "w")
+            # Use async writer
+            self._async_writer = AsyncLogWriter()
+            self._async_writer.start()
             
-        # Write config as first line
+            self._json_file_id = f"json_{timestamp}"
-        self._json_file.write(json.dumps({"type": "config", "data": config}) + "\n")
+            self._summary_file_id = f"summary_{timestamp}"
             
-        self._summary_file.write(f"Simulation Session Started: {datetime.now()}\n")
+            self._async_writer.open_file(str(json_path), self._json_file_id)
-        self._summary_file.write("=" * 60 + "\n\n")
+            self._async_writer.open_file(str(summary_path), self._summary_file_id)
+            
+            # Write initial data
+            self._async_writer.write(
+                self._json_file_id,
+                json.dumps({"type": "config", "data": config}) + "\n"
+            )
+            self._async_writer.write(
+                self._summary_file_id,
+                f"Simulation Session Started: {datetime.now()}\n" + "=" * 60 + "\n\n"
+            )
+        else:
+            # Use synchronous file handles
+            self._json_file = open(json_path, "w")
+            self._summary_file = open(summary_path, "w")
+            
+            self._json_file.write(json.dumps({"type": "config", "data": config}) + "\n")
+            self._summary_file.write(f"Simulation Session Started: {datetime.now()}\n")
+            self._summary_file.write("=" * 60 + "\n\n")
 
         if self.detailed_logging:
             # Set up file handler for detailed logs
@@ -263,36 +442,43 @@ class SimulationLogger:
         if not self.logging_enabled or self._current_entry is None:
             return
 
-        # Write to JSON lines file (without flush every time)
+        entry = self._current_entry
-        if self._json_file:
-            self._json_file.write(
-                json.dumps({"type": "turn", "data": self._current_entry.to_dict()}) + "\n"
-            )
         
-        # Write summary (without flush every time)
+        # Prepare data
-        if self._summary_file:
+        json_data = json.dumps({"type": "turn", "data": entry.to_dict()}) + "\n"
-            entry = self._current_entry
-            self._summary_file.write(
-                f"Turn {entry.turn} | Day {entry.day} Step {entry.step_in_day} ({entry.time_of_day})\n"
-            )
         
-            if self.detailed_logging:
+        summary_lines = [f"Turn {entry.turn} | Day {entry.day} Step {entry.step_in_day} ({entry.time_of_day})\n"]
-                for agent in entry.agent_entries:
-                    action = agent.decision.get("action", "?")
-                    result = "✓" if agent.action_result.get("success", False) else "✗"
-                    self._summary_file.write(
-                        f"  [{agent.agent_name}] {action} {result} | "
-                        f"E:{agent.stats_after.get('energy', '?')} "
-                        f"H:{agent.stats_after.get('hunger', '?')} "
-                        f"T:{agent.stats_after.get('thirst', '?')} "
-                        f"${agent.money_after}\n"
-                    )
         
-            if entry.deaths:
+        if self.detailed_logging:
-                for death in entry.deaths:
+            for agent in entry.agent_entries:
-                    self._summary_file.write(f"  💀 {death['name']} died: {death['cause']}\n")
+                action = agent.decision.get("action", "?")
+                result = "+" if agent.action_result.get("success", False) else "-"
+                summary_lines.append(
+                    f"  [{agent.agent_name}] {action} {result} | "
+                    f"E:{agent.stats_after.get('energy', '?')} "
+                    f"H:{agent.stats_after.get('hunger', '?')} "
+                    f"T:{agent.stats_after.get('thirst', '?')} "
+                    f"${agent.money_after}\n"
+                )
 
-            self._summary_file.write("\n")
+        if entry.deaths:
+            for death in entry.deaths:
+                summary_lines.append(f"  X {death['name']} died: {death['cause']}\n")
+
+        summary_lines.append("\n")
+        summary_data = "".join(summary_lines)
+        
+        if self.async_logging and self._async_writer:
+            # Non-blocking write
+            self._async_writer.write(self._json_file_id, json_data)
+            self._async_writer.write(self._summary_file_id, summary_data)
+            self._items_queued += 2
+        else:
+            # Synchronous write
+            if self._json_file:
+                self._json_file.write(json_data)
+            if self._summary_file:
+                self._summary_file.write(summary_data)
 
         # Batched flush - only flush every N turns
         self._turns_since_flush += 1
@@ -306,20 +492,41 @@ class SimulationLogger:
 
     def _flush_files(self) -> None:
         """Flush file buffers to disk."""
-        if self._json_file:
+        if self.async_logging and self._async_writer:
-            self._json_file.flush()
+            self._async_writer.flush()
-        if self._summary_file:
+        else:
-            self._summary_file.flush()
+            if self._json_file:
+                self._json_file.flush()
+            if self._summary_file:
+                self._summary_file.flush()
 
     def close(self) -> None:
         """Close log files."""
-        if self._json_file:
+        if self.async_logging and self._async_writer:
-            self._json_file.close()
+            # Write final message before closing
-            self._json_file = None
+            if self._summary_file_id:
-        if self._summary_file:
+                self._async_writer.write(
-            self._summary_file.write(f"\nSession ended: {datetime.now()}\n")
+                    self._summary_file_id,
-            self._summary_file.close()
+                    f"\nSession ended: {datetime.now()}\n"
-            self._summary_file = None
+                )
+            
+            # Close files
+            if self._json_file_id:
+                self._async_writer.close_file(self._json_file_id)
+            if self._summary_file_id:
+                self._async_writer.close_file(self._summary_file_id)
+            
+            # Stop the writer thread
+            self._async_writer.stop()
+            self._async_writer = None
+        else:
+            if self._json_file:
+                self._json_file.close()
+                self._json_file = None
+            if self._summary_file:
+                self._summary_file.write(f"\nSession ended: {datetime.now()}\n")
+                self._summary_file.close()
+                self._summary_file = None
 
     def get_entries(self) -> list[TurnLogEntry]:
         """Get all logged entries.
@@ -328,6 +535,15 @@ class SimulationLogger:
         """
         return []
     
+    def get_stats(self) -> dict:
+        """Get logging statistics."""
+        return {
+            "logging_enabled": self.logging_enabled,
+            "async_logging": self.async_logging,
+            "items_queued": self._items_queued,
+            "items_dropped": self._items_dropped,
+        }
+
 
 # Global logger instance
 _logger: Optional[SimulationLogger] = None
@@ -348,3 +564,12 @@ def reset_simulation_logger() -> SimulationLogger:
         _logger.close()
     _logger = SimulationLogger()
     return _logger
+
+
+# Ensure logger is closed on exit
+@atexit.register
+def _cleanup_logger():
+    global _logger
+    if _logger:
+        _logger.close()
+        _logger = None

backend/core/storage.py

@@ -0,0 +1,450 @@
+"""State storage abstraction for simulation state.
+
+Provides a unified interface for storing and retrieving simulation state,
+with implementations for:
+- In-memory storage (default, fast but not persistent)
+- Redis storage (optional, enables decoupled UI polling and persistence)
+
+This allows the simulation loop to snapshot state without blocking,
+and enables external systems (like a web UI) to poll state independently.
+"""
+
+import json
+import time
+from abc import ABC, abstractmethod
+from dataclasses import dataclass
+from typing import Optional, Dict
+
+
+@dataclass
+class StateSnapshot:
+    """A point-in-time snapshot of simulation state."""
+    turn: int
+    timestamp: float
+    data: dict
+    
+    def to_json(self) -> str:
+        """Convert to JSON string."""
+        return json.dumps({
+            "turn": self.turn,
+            "timestamp": self.timestamp,
+            "data": self.data,
+        })
+    
+    @classmethod
+    def from_json(cls, json_str: str) -> "StateSnapshot":
+        """Create from JSON string."""
+        obj = json.loads(json_str)
+        return cls(
+            turn=obj["turn"],
+            timestamp=obj["timestamp"],
+            data=obj["data"],
+        )
+
+
+class StateStore(ABC):
+    """Abstract interface for state storage.
+    
+    Implementations should be thread-safe for concurrent read/write.
+    """
+    
+    @abstractmethod
+    def save_state(self, key: str, snapshot: StateSnapshot) -> bool:
+        """Save a state snapshot.
+        
+        Args:
+            key: Unique key for this state (e.g., "world", "market", "agent_123")
+            snapshot: The state snapshot to save
+            
+        Returns:
+            True if saved successfully
+        """
+        pass
+    
+    @abstractmethod
+    def get_state(self, key: str) -> Optional[StateSnapshot]:
+        """Get the latest state snapshot for a key.
+        
+        Args:
+            key: The state key to retrieve
+            
+        Returns:
+            The snapshot if found, None otherwise
+        """
+        pass
+    
+    @abstractmethod
+    def get_all_states(self, prefix: str = "") -> Dict[str, StateSnapshot]:
+        """Get all states matching a prefix.
+        
+        Args:
+            prefix: Key prefix to filter by (empty = all)
+            
+        Returns:
+            Dict mapping keys to snapshots
+        """
+        pass
+    
+    @abstractmethod
+    def delete_state(self, key: str) -> bool:
+        """Delete a state snapshot.
+        
+        Args:
+            key: The state key to delete
+            
+        Returns:
+            True if deleted (or didn't exist)
+        """
+        pass
+    
+    @abstractmethod
+    def clear_all(self) -> None:
+        """Clear all stored states."""
+        pass
+    
+    @abstractmethod
+    def is_healthy(self) -> bool:
+        """Check if the store is healthy and accessible."""
+        pass
+
+
+class MemoryStateStore(StateStore):
+    """In-memory state storage (default implementation).
+    
+    Fast but not persistent across restarts.
+    Thread-safe using a simple lock.
+    """
+    
+    def __init__(self, max_entries: int = 1000):
+        """Initialize memory store.
+        
+        Args:
+            max_entries: Maximum number of entries to keep (LRU eviction)
+        """
+        import threading
+        self._data: Dict[str, StateSnapshot] = {}
+        self._lock = threading.Lock()
+        self._max_entries = max_entries
+        self._access_order: list[str] = []  # For LRU tracking
+    
+    def save_state(self, key: str, snapshot: StateSnapshot) -> bool:
+        with self._lock:
+            # LRU eviction if at capacity
+            if len(self._data) >= self._max_entries and key not in self._data:
+                oldest = self._access_order.pop(0) if self._access_order else None
+                if oldest:
+                    self._data.pop(oldest, None)
+            
+            self._data[key] = snapshot
+            
+            # Update access order
+            if key in self._access_order:
+                self._access_order.remove(key)
+            self._access_order.append(key)
+            
+            return True
+    
+    def get_state(self, key: str) -> Optional[StateSnapshot]:
+        with self._lock:
+            snapshot = self._data.get(key)
+            if snapshot and key in self._access_order:
+                # Update access order for LRU
+                self._access_order.remove(key)
+                self._access_order.append(key)
+            return snapshot
+    
+    def get_all_states(self, prefix: str = "") -> Dict[str, StateSnapshot]:
+        with self._lock:
+            if not prefix:
+                return dict(self._data)
+            return {k: v for k, v in self._data.items() if k.startswith(prefix)}
+    
+    def delete_state(self, key: str) -> bool:
+        with self._lock:
+            self._data.pop(key, None)
+            if key in self._access_order:
+                self._access_order.remove(key)
+            return True
+    
+    def clear_all(self) -> None:
+        with self._lock:
+            self._data.clear()
+            self._access_order.clear()
+    
+    def is_healthy(self) -> bool:
+        return True
+
+
+class RedisStateStore(StateStore):
+    """Redis-backed state storage.
+    
+    Enables:
+    - Persistent state across restarts
+    - Decoupled UI polling (web clients can read state independently)
+    - Distributed access (multiple simulation instances)
+    
+    Requires redis-py: pip install redis
+    """
+    
+    def __init__(
+        self,
+        host: str = "localhost",
+        port: int = 6379,
+        db: int = 0,
+        password: Optional[str] = None,
+        prefix: str = "villsim:",
+        ttl_seconds: int = 3600,  # 1 hour default TTL
+    ):
+        """Initialize Redis store.
+        
+        Args:
+            host: Redis server host
+            port: Redis server port
+            db: Redis database number
+            password: Redis password (if required)
+            prefix: Key prefix for all keys (for namespacing)
+            ttl_seconds: Time-to-live for entries (0 = no expiry)
+        """
+        self._prefix = prefix
+        self._ttl = ttl_seconds
+        self._client = None
+        self._connection_params = {
+            "host": host,
+            "port": port,
+            "db": db,
+            "password": password,
+            "decode_responses": True,
+        }
+        self._connect()
+    
+    def _connect(self) -> None:
+        """Establish connection to Redis."""
+        try:
+            import redis
+            self._client = redis.Redis(**self._connection_params)
+            # Test connection
+            self._client.ping()
+        except ImportError:
+            raise ImportError(
+                "Redis support requires the 'redis' package. "
+                "Install with: pip install redis"
+            )
+        except Exception as e:
+            self._client = None
+            raise ConnectionError(f"Failed to connect to Redis: {e}")
+    
+    def _make_key(self, key: str) -> str:
+        """Create full Redis key with prefix."""
+        return f"{self._prefix}{key}"
+    
+    def save_state(self, key: str, snapshot: StateSnapshot) -> bool:
+        if not self._client:
+            return False
+        
+        try:
+            full_key = self._make_key(key)
+            data = snapshot.to_json()
+            
+            if self._ttl > 0:
+                self._client.setex(full_key, self._ttl, data)
+            else:
+                self._client.set(full_key, data)
+            
+            return True
+        except Exception:
+            return False
+    
+    def get_state(self, key: str) -> Optional[StateSnapshot]:
+        if not self._client:
+            return None
+        
+        try:
+            full_key = self._make_key(key)
+            data = self._client.get(full_key)
+            
+            if data:
+                return StateSnapshot.from_json(data)
+            return None
+        except Exception:
+            return None
+    
+    def get_all_states(self, prefix: str = "") -> Dict[str, StateSnapshot]:
+        if not self._client:
+            return {}
+        
+        try:
+            pattern = self._make_key(prefix + "*")
+            keys = self._client.keys(pattern)
+            
+            result = {}
+            for full_key in keys:
+                # Remove our prefix to get the original key
+                key = full_key[len(self._prefix):]
+                data = self._client.get(full_key)
+                if data:
+                    result[key] = StateSnapshot.from_json(data)
+            
+            return result
+        except Exception:
+            return {}
+    
+    def delete_state(self, key: str) -> bool:
+        if not self._client:
+            return False
+        
+        try:
+            full_key = self._make_key(key)
+            self._client.delete(full_key)
+            return True
+        except Exception:
+            return False
+    
+    def clear_all(self) -> None:
+        if not self._client:
+            return
+        
+        try:
+            pattern = self._make_key("*")
+            keys = self._client.keys(pattern)
+            if keys:
+                self._client.delete(*keys)
+        except Exception:
+            pass
+    
+    def is_healthy(self) -> bool:
+        if not self._client:
+            return False
+        
+        try:
+            self._client.ping()
+            return True
+        except Exception:
+            return False
+    
+    def publish_state_update(self, channel: str, key: str) -> None:
+        """Publish a state update notification (for real-time subscribers).
+        
+        This can be used for WebSocket-style updates where clients
+        subscribe to state changes.
+        """
+        if not self._client:
+            return
+        
+        try:
+            self._client.publish(
+                f"{self._prefix}updates:{channel}",
+                json.dumps({"key": key, "timestamp": time.time()})
+            )
+        except Exception:
+            pass
+
+
+class StubStateStore(StateStore):
+    """No-op state store for when storage is disabled.
+    
+    All operations succeed but don't actually store anything.
+    """
+    
+    def save_state(self, key: str, snapshot: StateSnapshot) -> bool:
+        return True
+    
+    def get_state(self, key: str) -> Optional[StateSnapshot]:
+        return None
+    
+    def get_all_states(self, prefix: str = "") -> Dict[str, StateSnapshot]:
+        return {}
+    
+    def delete_state(self, key: str) -> bool:
+        return True
+    
+    def clear_all(self) -> None:
+        pass
+    
+    def is_healthy(self) -> bool:
+        return True
+
+
+# Global state store instance
+_state_store: Optional[StateStore] = None
+
+
+def get_state_store() -> StateStore:
+    """Get the global state store instance.
+    
+    Creates a store based on config:
+    - If Redis is configured and available, uses Redis
+    - Otherwise falls back to in-memory storage
+    """
+    global _state_store
+    
+    if _state_store is None:
+        _state_store = _create_state_store()
+    
+    return _state_store
+
+
+def _create_state_store() -> StateStore:
+    """Create the appropriate state store based on config."""
+    from backend.config import get_config
+    
+    config = get_config()
+    
+    # Check for Redis config
+    redis_config = getattr(config, 'redis', None)
+    
+    if redis_config and getattr(redis_config, 'enabled', False):
+        try:
+            store = RedisStateStore(
+                host=getattr(redis_config, 'host', 'localhost'),
+                port=getattr(redis_config, 'port', 6379),
+                db=getattr(redis_config, 'db', 0),
+                password=getattr(redis_config, 'password', None),
+                prefix=getattr(redis_config, 'prefix', 'villsim:'),
+                ttl_seconds=getattr(redis_config, 'ttl_seconds', 3600),
+            )
+            if store.is_healthy():
+                return store
+        except Exception:
+            # Fall through to memory store
+            pass
+    
+    # Check if storage is disabled
+    perf_config = getattr(config, 'performance', None)
+    if perf_config and not getattr(perf_config, 'state_storage_enabled', True):
+        return StubStateStore()
+    
+    # Default to memory store
+    return MemoryStateStore()
+
+
+def reset_state_store() -> None:
+    """Reset the global state store."""
+    global _state_store
+    if _state_store:
+        _state_store.clear_all()
+    _state_store = None
+
+
+def save_simulation_state(turn: int, state_data: dict) -> bool:
+    """Convenience function to save simulation state.
+    
+    Args:
+        turn: Current simulation turn
+        state_data: Full state dict (world, market, agents, etc.)
+        
+    Returns:
+        True if saved successfully
+    """
+    store = get_state_store()
+    snapshot = StateSnapshot(
+        turn=turn,
+        timestamp=time.time(),
+        data=state_data,
+    )
+    return store.save_state("simulation:current", snapshot)
+
+
+def get_simulation_state() -> Optional[StateSnapshot]:
+    """Convenience function to get current simulation state."""
+    store = get_state_store()
+    return store.get_state("simulation:current")

config.json

@@ -2,14 +2,33 @@
   "performance": {
     "logging_enabled": false,
     "detailed_logging": false,
+    "async_logging": true,
     "log_flush_interval": 50,
     "max_turn_logs": 100,
-    "stats_update_interval": 10
+    "stats_update_interval": 10,
+    "state_storage_enabled": true
   },
   "ai": {
     "goap_max_iterations": 30,
     "goap_max_plan_depth": 2,
-    "reactive_fallback": true
+    "reactive_fallback": true,
+    "use_bdi": true
+  },
+  "bdi": {
+    "thinking_interval": 1,
+    "max_consecutive_failures": 2,
+    "priority_switch_threshold": 1.5,
+    "memory_max_events": 50,
+    "memory_decay_rate": 0.1
+  },
+  "redis": {
+    "enabled": false,
+    "host": "localhost",
+    "port": 6379,
+    "db": 0,
+    "password": null,
+    "prefix": "villsim:",
+    "ttl_seconds": 3600
   },
   "agent_stats": {
     "max_energy": 50,