deeplethe/utopia

World's first open-source enterprise world model.

What it solves

Utopia is a self-hosted knowledge platform designed to track not just what information is known, but when it was true. It solves the problem of outdated or changing information in organizational knowledge bases by treating facts as append-only and assigning validity intervals to every piece of data, allowing users to replay history and see the state of knowledge at any specific point in time.

How it works

The system uses a hybrid retrieval approach combining full-text search (Tantivy) and vector embeddings (pgvector) via reciprocal rank fusion. It extracts entities and facts from uploaded documents (PDF, DOCX, Markdown, etc.) using an LLM based on a customizable ontology. These facts are stored in a PostgreSQL database as a knowledge graph where every edge links back to the source sentence for evidence. It also includes a review queue for human verification of low-confidence extractions.

Who it’s for

It is built for organizations that need a secure, air-gapped or self-hosted RAG system that maintains a strict audit trail of how knowledge evolves over time.

Highlights

  • Temporal Knowledge Graph: A timeline-based graph that allows users to visualize and and query facts as they existed at a specific date.
  • Hybrid Search: Combines keyword and vector search for more accurate retrieval with inline citations.
  • Simplified Infrastructure: Runs as a single binary and one Postgres database, removing the need for separate vector services or message queues.
  • Multi-tenant Support: Includes role-based access control and organizational workspaces for managing multiple knowledge bases.
  • Database Integration: Ability to mount external Postgres connections to query structured data alongside unstructured documents.

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
    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
    trustgraph-ai/trustgraph

    An open-source context orchestration layer that uses hypergraphs to turn raw enterprise data into structured, deterministic AI context for agentic AI.

  • 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
    xerrors/Yuxi

    A self-deployable, multi-tenant knowledge agent platform that integrates RAG, knowledge graphs, and multi-agent orchestration for secure team collaboration.