deeplethe/utopia

World's first open-source enterprise world model.

What it solves

Utopia addresses the limitations of standard vector stores and knowledge graphs by introducing time-awareness and formal ontology into the base layer of enterprise knowledge management. It prevents the loss of historical context by recording how understanding evolves over time, allowing companies to maintain a trustworthy decision core and a full audit trail of how facts were derived or corrected.

How it works

Built with Rust and Postgres (using pgvector), Utopia implements a bitemporal knowledge graph. It extracts entities and facts from various documents and data sources, mapping them to an ontology. The system tracks two timelines: when a fact was true in the real world and when the system recorded it. It uses a combination of full-text search (Tantivy) and vector search (RRF fusion) to provide cited answers via LLMs. It also employs forward-chaining reasoning to derive new facts from ontology axioms and includes a conflict detection system to handle contradictory data.

Who it’s for

It is designed for enterprises that need a high-integrity knowledge foundation for AI agents, specifically those requiring strict compliance, auditability, and the ability to reason over complex, changing business data across documents and databases.

Highlights

  • Bitemporal Knowledge Graph: Tracks both the validity time of a fact and the system's time of recording, ensuring a complete history of knowledge changes.
  • Ontology-Driven RAG: Integrates structured ontology and the Ontology2SQL method to query both unstructured documents and mounted databases (Postgres, MySQL, Snowflake, etc.).
  • Agentic Capabilities: Includes a built-in agent that can walk the knowledge graph, search documents, and query databases, also exposed via MCP.
  • Self-Governing Logic: Uses ontology axioms to automatically derive new facts and detect logical conflicts in the data.
  • Air-Gapped Deployment: Supports any OpenAI-compatible endpoint, allowing the entire system to run on private hardware.

Related

  • Project
    getzep/graphiti

    Graphiti is a framework for building temporal knowledge graphs that track how facts change over time, enabling AI agents to maintain evolving, high-provenance context.

  • Project
    Ais1on/CTI-RAG

    A RAG system specialized for Cyber Threat Intelligence that combines vector search and knowledge graphs to enable multi-hop analysis of attack organizations and malware.

  • Project
    HKUDS/LightRAG

    A lightweight, graph-based RAG framework that combines knowledge graphs and vector embeddings to provide efficient, high-quality retrieval and global context understanding.

  • Project
    xoai/sage-wiki

    A graph-based knowledge base that uses LLMs to compile raw documents into an interlinked wiki and knowledge graph for shared human-AI memory.

  • Project
    Zleap-AI/SAG

    A knowledge base application using a novel SAG architecture that replaces traditional RAG and GraphRAG with query-time dynamic hyperedges for superior retrieval and relational reasoning.