wyrd-ecs-core

WYRD ECS Core is an experimental open-source framework that implements a world model using an Entity-Component-System (ECS) architecture. It reimagines how AI agents consume structured, real-time data by moving world modeling from ephemeral LLM memory into a persistent, physicalized digital universe. Every entity carries explicit relationships, vectors, and timestamps, making it a living data ecosystem rather than static snapshots. Built for developers building cognitive infrastructures, autonomous agents, or simulation environments, WYRD ECS Core offers dynamic typing that lets components emerge from data patterns, a state-driven component stream for continuous telemetry, a system orchestrator handling millions of updates per second, and real-time data federation from diverse sources like IoT sensors and financial markets. The project also includes temporal synchronization for latency mismatch and fault-tolerant ingestion with ghost components. The framework supports AI world model synthesis using Graph Neural Networks and Transformers, with state space awareness that lets models navigate graphs of real-world entities directly. It tracks causality automatically, registers causal links between component changes, and can generate probabilistic future states from partial world data without hallucination. Additionally, it employs a sovereign data architecture with distributed hash graph storage, avoiding central databases. Positioned as a conceptual prototype rather than a production-ready tool, WYRD ECS Core differs from alternatives like LangChain or AutoGPT by focusing on structured world representation instead of prompt chains. It is best suited for researchers and advanced developers who want to experiment with ECS-based architectures for AI cognition.