AI-Driven Open World NPC System

You are an AI systems architect for open-world games. Design a complete AI-driven NPC system: Game genre: [INSERT - open world RPG, survival, simulation] Engine: [INSERT - Unreal Engine 5, Unity, custom] LLM backend: [INSERT - local model, cloud API, fine-tuned] NPC count target: [INSERT] Cover the full pipeline: 1. NPC core architecture - entity component structure, memory store design (episodic, semantic, procedural), belief-desire-intention model, personality traits and values. 2. LLM integration layer - prompt templates for NPC dialogue and decisions, token budget management, response parsing, hallucination guards, offline fallback rule-based AI. 3. Memory and persistence - short-term and long-term memory, memory consolidation and summarization, memory salience and decay, cross-session persistence, save/load. 4. Social relationship graph - NPC-to-NPC affinity, factions, reputation, relationship state machine, gossip and information spread, conflict and alliance formation. 5. Dynamic schedules and goals - goal generation from needs hierarchy, daily/weekly routines, adaptive behavior to world events, emergent cooperation and competition. 6. Dialogue system - context-aware conversation, player choice branching, NPC voice and tone consistency, dialogue state tracking, quest hooks through dialogue. 7. Performance and scaling - LLM request batching and caching, simulated vs active NPCs, LOD for AI complexity, budget for concurrent LLM calls, streaming response.