Utopia: The Open-Source Enterprise World Model for Knowledge Engineering

Introduction

Utopia, developed by DeepLethe, is an innovative open-source enterprise world model written in Rust. Unlike traditional knowledge graphs or vector stores that focus on present knowledge, Utopia integrates time awareness and ontology at its core. This system passively learns, governs itself, and allows knowledge to evolve as new material arrives, ensuring conflict detection, reasoning, and decision-making are aligned with a dynamic ontology. It is designed for offline deployment, providing a secure, self-hosted solution for building a trustworthy knowledge foundation and decision core for enterprise agents.

Why Use and Key Benefits

Utopia offers significant advantages for organizations seeking advanced knowledge management and decision intelligence. Its bitemporal knowledge graph records the entire history of changing understanding, allowing for comprehensive review of past decisions with full context and grounds. Key features include:

A Complete Application: A single Rust binary and Postgres setup provides a system console, graph browser, and ontology workbench in one web UI.
Extensive Knowledge Ingest: Supports various document formats (PDF, DOCX, Markdown, HTML, etc.) and syncs from web pages, RSS, GitHub, Jira, Notion, and S3-compatible buckets.
Advanced Search and Chat: Combines full-text search (Tantivy) with vector search (pgvector) using RRF, offering streaming answers with inline citations. It supports any OpenAI-compatible endpoint, enabling air-gapped operation.
Agent Harness and Agentic RAG: The system can be driven conversationally, allowing agents to search documents, traverse the graph, and query mounted databases.
Ontology-Driven Intelligence: Starts with pre-defined ontology packs (schema.org, W3C Org, PROV-O, FOAF, IOF Core) and allows for custom vocabulary evolution.
Bitemporal Graph: Facts carry temporal validity and origin, preserving both when something was true and when the system believed it.
Conflict Detection and Reasoning: Ontology axioms compile into rules for deriving new facts and detecting conflicts, ensuring data integrity.
Ontology-driven Querying: Mounts databases (Postgres, MySQL, Trino, Databricks, Snowflake) for chat-based querying alongside documents, powered by state-of-the-art Ontology2SQL.
Decision Ledger: Maintains an append-only record of all system decisions, providing a robust audit trail.

Installation

To get started with Utopia, you will need Docker.

Clone the repository:

git clone https://github.com/deeplethe/utopia.git
cd utopia

Start from the prebuilt image:
```
docker compose --profile app up -d
```
Access the application at http://localhost:1516 and register. Remember to configure model endpoints (chat and embedding) under Administration ? Models before extracting documents.

For local development, you will also need Rust 1.85+, Node 20+, and pnpm. Refer to the official GitHub README for detailed instructions.

Examples

Utopia empowers users to:

Ingest and Organize Knowledge: Upload diverse documents and automatically extract entities and facts, building a structured knowledge base.
Perform Semantic Search and Chat: Query the knowledge base using natural language, receiving precise answers with citations, even in air-gapped environments.
Drive AI Agents: Utilize the built-in agent or connect external agent frameworks via MCP to automate tasks, explore historical data, and make informed decisions.
Manage Ontologies: Define and evolve the vocabulary of your knowledge base, ensuring consistency and accuracy across all data.
Review and Audit Decisions: Leverage the bitemporal graph and decision ledger to understand the evolution of knowledge and the rationale behind past decisions.
Integrate with Databases: Connect to existing databases and query them alongside documents using ontology-driven text-to-SQL capabilities.