Self-host Utopia
Bitemporal knowledge graph with ontology-driven RAG
- Stars
- 8 585
- Port
- 1516 (Web UI and API served by the Utopia app container)
- Requires
- PostgreSQL with pgvector
ghcr.io/deeplethe/utopiaWhat is Utopia?
Utopia is an open-source knowledge foundation for enterprises from DeepLethe, pitched as an "enterprise world model" rather than just another vector store. Time and ontology sit in its base layer: every stored fact notes the period it held and the document it came from, and a correction seals off the earlier statement instead of erasing it, leaving the graph with two parallel timelines (bitemporal). Ingest accepts PDFs, Word, PowerPoint, spreadsheets, Markdown, HTML and plain text; scans and pictures go to MinerU or a vision model from Volcengine Ark, and recordings are transcribed with speaker labels. Scheduled connectors bring in web pages, RSS feeds, S3-compatible buckets, plus Jira, Notion, GitHub and WebDAV sources. Retrieval fuses Tantivy full-text with pgvector embeddings through RRF; chat answers stream with clickable citations, and model endpoints only need to be OpenAI-compatible (DeepSeek, Qwen, GLM, Ollama, vLLM), which makes fully offline, air-gapped operation possible. A built-in agent lets you converse with the whole system, and an MCP server exposes its read-only tools to frameworks like Claude Desktop and Cursor. New bases start from bundled ontology packs (the W3C Org, FOAF, PROV-O, IOF Core and schema.org sets); axioms compile into forward-chaining derivation rules that stay off by default, while doubtful or clashing extractions land in a human review queue. Mounted external databases — Postgres, MySQL, ClickHouse, Trino, Databricks or Snowflake — can be queried in chat alongside the documents. Bases carry owner/admin/editor/viewer roles, and every confirm, merge or rebuild is logged in an append-only decision ledger. Self-hosting means one Docker Compose stack: a single Rust binary plus a Postgres container with pgvector, with the web UI on port 1516. Caveats: the project is still v0.1, migrations roll forward only with no rollback, published images target linux/amd64 and linux/arm64, and decision replay/reasoning remains on the roadmap.
Should you self-host Utopia?
Utopia scores 57/100 on selfhost.directory (how scoring works) and ranks #15 of 42 in Knowledge Management. It has 8 585 GitHub stars.
There was no stable release in the past 12 months, so check the repository's recent activity before you rely on it.
Setup is moderate: expect some configuration beyond a single command. An official Docker image is available. It is Apache-2.0-licensed and written mostly in Rust.
Its closest alternative on selfhost.directory is Project N.O.M.A.D. (85/100, 41 116 stars). It scores higher overall, so compare the two before you pick. See Utopia vs Project N.O.M.A.D. side by side.
Features
- Agentic RAG
- Bitemporal knowledge graph
- Chat with inline citations
- Conflict Detection
- Document Ingestion
- Entity Resolution
- Hybrid full-text and vector search
- MCP Server
- Ontology-driven database querying
- Ontology packs
- Review Queue
- Roles and append-only decision ledger
- Rule-based derivation from ontology axiomsDerivation is off by default.
- Scheduled source sync
- Web UI with console, graph browser and ontology workbench
Utopia README
More in Graph Database
arangodbSelf-hosted multi-model database for documents, graphs and key-values, queried in one AQL language; the open-source Community edition is free for non-commercial use.
ArcadeDBOpen-source, self-hosted multi-model database: one ACID engine for graph, document, key-value, vector, full-text and time-series data, queried in SQL, Cypher or Gremlin.
FalkorDBA super fast Graph Database uses GraphBLAS under the hood for its sparse adjacency matrix graph representation. Our goal is to provide the best Knowledge Graph for LLM (GraphRAG).
GelOpen-source graph-relational database that builds on Postgres with EdgeQL, migrations, branching, built-in auth and AI; self-hosted via Docker or the gel-server binary.
gunOpen-source P2P graph database with realtime sync and offline-first storage; run a self-hosted relay peer via Docker or npm in browsers, Node.js and mobile apps.
helixdbA Rust-based graph database that unifies vector similarity and keyword search on object storage for AI memory workloads.
memgraphSelf-hosted, open-source in-memory graph database with Cypher, vector indexes and the MAGE algorithm library for GraphRAG and real-time graph analytics.
neo4jSelf-hosted, open-source graph database: query connected data with Cypher under ACID transactions. Run the free GPLv3 Community Edition via Docker, packages, or Helm.
rushdbRushDB is a self-hosted graph+vector DB for AI agents unifying relationships, semantic search, and live schema in one API, replacing Redis+Pinecone+Neo4j.
SurrealDBSelf-hosted, source-available Rust database unifying documents, graphs, vectors and time-series, with live queries and row-level permissions in SurrealQL.
0 comments