Verified project record

deeplethe/utopia

Utopia is a single Rust binary plus Postgres that builds an ontology-governed bitemporal knowledge graph from documents and databases, offering search, chat with citations, agentic RAG, and MCP integration, with the ability to run air-gapped. It targets enterprises that need time-aware knowledge with provenance for decisions and compliance rather than just present-state storage.

Project overview

It records when each fact held and its source, closing old versions instead of overwriting them, and every knowledge base exposes an MCP server for agent frameworks to connect.

Project type: MCP · RAG · AI Search
Use cases: Knowledge Q&A · Search & Research · Data Analysis
Deployment: Refer to project documentation
License: Apache-2.0

Best for

Enterprise teams that need a time-aware, ontology-governed knowledge foundation with source provenance and an offline deployment path for decisions and compliance.
Data teams and AI engineers who want documents and mounted databases (Postgres, MySQL, Trino, Databricks, Snowflake) queryable together through chat with citations and per-knowledge-base MCP servers.

Key capabilities

System console, graph browser and ontology workbench in one web UI; install and it works.
Upload PDF, DOCX, PPTX, XLSX, XLS, ODS, CSV, TSV, Markdown, HTML or plain text; scheduled sync of web pages, RSS, GitHub, Jira, Notion, WebDAV and S3-compatible buckets; an API covers everything else.
Full-text search on Tantivy fused with pgvector vectors via RRF; streaming answers with inline citations; works with any OpenAI-compatible endpoint so the system can run air-gapped.
Every knowledge base exposes an MCP server with fine-grained permissions so agent frameworks can connect.
Extraction turns documents into entities and facts; every fact carries when it held and source; corrections close old versions rather than overwriting; edges are reified.
Mount Postgres, MySQL, Trino, Databricks, Snowflake databases and chat can query them alongside documents via Ontology2SQL mapping.

Limitations and risks

Still at v0.1; the database schema evolves between versions and migrations only roll forward with no rollback.
Derivation is disabled by default because a wrong axiom derives wrong facts; expect to enable and review it deliberately.

Getting started

Setup is straightforward: clone the repository, run 'docker compose --profile app up -d', then open http://localhost:1516 and register; the first account automatically becomes administrator. OpenAI-compatible chat and embedding model endpoints must then be configured under Administration → Models. No coding is required.

Alternatives and comparisons

Consider when the need is search over local files for AI agents rather than an ontology-governed bitemporal enterprise knowledge graph.
Consider when the need is question-answering over local documents rather than database mounts, ontology governance, and temporal provenance.
Consider when the need is document retrieval within a coding assistant rather than a self-hosted bitemporal knowledge foundation with chat citations and MCP per knowledge base.

Project comparisons

Evidence and sources

README: One Rust binary and one Postgres. Full-text search is embedded in the binary, vectors go in pgvector, and the job queue is a table: nothing else to run.
README: It deploys offline, so a company can stand up a knowledge foundation, a decision core its agents can trust, and a compliance audit trail on hardware it controls.
README: Where existing vector stores and knowledge graphs work to get present knowledge right, one of Utopia's founding aims is to record the whole course of changing understanding. Engin…
README: A system console, a graph browser and an ontology workbench in one web UI. A product, not a library: install it and it works.
README: Upload PDF, DOCX, PPTX, XLSX, XLS, ODS, CSV, TSV, Markdown, HTML or plain text, with legacy encodings detected on the way in. Web pages, RSS, GitHub, Jira, Notion, WebDAV and S3-c…

Source updated at: 2026-09-22T19:00:15Z

Open original GitHub project