Utopia: Open-Source Enterprise World Model with Bitemporal Knowledge Graph

Utopia is an open-source enterprise world model that integrates knowledge graphs, vector retrieval, ontology management, and AI reasoning into a privately deployable product. Its core innovation is a bitemporal knowledge graph recording both world time (when a fact was true) and belief time (when the system believed it), preserving history for audit and decision intelligence. The pipeline includes document parsing, entity extraction with signature checks and 11 discard reason codes, three-stage entity disambiguation (exact name, embedding similarity, model adjudication) with thresholds 0.55 merge and 0.35 create, type resolution via extraction clues and neighbor voting, ontology growth from OOV proposals, and consistency checking with materialized inference (transitive, symmetric, inverse) stored in a separate derived_facts table. The project includes 26 architecture decision records (ADRs) covering design choices such as 'no type is a type', contradictions as data quality signals, and support for RDF export to avoid vendor lock-in.