Abstract

An Agentic Knowledge Graph Engine (AKGE) is disclosed that provides a complete, continuously evolving knowledge lifecycle managed by a multi-agent autonomous architecture. The system comprises a data extraction layer receiving heterogeneous input sources, a schema-enforced knowledge graph storage layer with native temporal versioning and audit lineage, a hybrid embedding module fusing semantic, structural, contextual, and schema-constrained representations into dynamically updated entity embeddings, a continuous evolution engine responding to event triggers including query failure, embedding drift, conflict detection, and new data arrival, and a hybrid retrieval engine combining multi-hop graph traversal with approximate nearest-neighbor vector search. A specialized multi-agent ensemble — comprising Extraction, Creation, Validation, Enrichment, Inference, Feedback, and Pruning agents — coordinates all lifecycle phases under defined graph mutation authorities and a shared ontological schema with hierarchical inheritance and selective overrides. Retrieval outcome signals are fed back to the Feedback Agent, which drives structural mutations, embedding fusion weight updates, and retrieval strategy refinements, forming a closed-loop self-improving knowledge system. The disclosure contains full end-to-end agentic lifecycle management; runtime continuous evolution (not batch or training-time); hybrid multi-signal embedding fusion with incremental re-computation; event-driven feedback-to-mutation loop; and native temporal versioning with knowledge lineage.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Share Alike 4.0 License.

Download

Share

COinS