The same small dataset and the same four steps on 12 open-source graph databases, each in a fresh container at a pinned version. Every command and its real output is saved in output/.
This is the test behind Best open source graph databases on omnigraph.dev. It's written by the team behind Omnigraph, one of the engines compared, which is why everything here is a script you can rerun rather than a claim.
Dataset (dataset.json): 12 people (name, age), 3 companies, 5 projects, WorksAt and ContributesTo edges, and three probe records.
- Schema and load. The schema each engine's docs recommend for typed data, then the full dataset.
- Two-hop query. Which projects do Acme's employees contribute to? Checked against the answer computed from the dataset (
Atlas, Beacon, Comet). - Schema probes. A person with
age: "forty", then one with an undeclarednickname, through the engine's normal write path. Rejected or accepted, in the default setup and again in the engine's strict mode when it documents one. - Branch, merge and a historical read. Branch, write on it, check main is untouched, merge, then read the commit from before. Run only where the engine documents branches or versioned reads; elsewhere recorded as
not available.
| Engine (version run) | Two-hop query | age: "forty" |
undeclared nickname |
Strict mode | Branch, merge, older commit |
|---|---|---|---|---|---|
| Omnigraph v0.15.0 | pass | rejected | rejected | not needed | pass, pass, pass |
| TerminusDB v12.0.9 | pass | rejected | rejected | not needed | pass, pass, pass |
| TypeDB CE 3.14.2 | pass | rejected | rejected | not needed | not available |
| LadybugDB 0.22.3 | pass | rejected | rejected | not needed | not available |
| NebulaGraph 3.8.0 | pass | rejected | rejected | not needed | not available |
| Dgraph v25.5.1 | pass | rejected | accepted | rejects both | not available |
| JanusGraph 1.2.0 | pass | rejected | accepted | rejects both | not available |
| OrientDB 3.2.58 | pass | rejected | accepted | rejects both | not available |
| ArcadeDB 26.11.1 | pass | rejected | accepted | none documented | not available |
| Neo4j CE 2026.10.2 | pass | accepted | accepted | refused in Community Edition | not available |
| Apache AGE 1.8.0 (PostgreSQL 18) | pass | accepted | accepted | none documented | not available |
| HelixDB (server image v0.0.15) | pass | accepted | accepted | none documented | not available |
Full logs: output/<engine>.txt. Machine-readable: output/results.json.
Requirements: Python 3.11+, Docker, uv (for LadybugDB), and the omnigraph CLI (OMNIGRAPH_BIN overrides its path).
python run_all.py # every engine
python run_all.py neo4j # one engineEach runner starts its own containers (named cmp-<engine>) and removes them when it's done. Images and package versions are pinned at the top of each <engine>/run.py and named in the first line of its log. Logs replace machine paths with $WORK (a temp dir), $CODE (this repo) and ~.
- TerminusDB:
works_atis optional in the schema, since the probe records have no company; the merge usesterminusdb applywith the branch's before and after commits. - NebulaGraph: the console exits 0 on errors, so the runner reads its output for
[ERROR. - HelixDB: v3 has no schema file or declared types, so step 1 only loads data. The query runs through the native DSL and is cross-checked through
/v2/cypher. - OrientDB: the console takes no parameters, so
'acme'is inlined. - JanusGraph: BerkeleyDB JE with Lucene, the single-container setup. Strict mode is a second container with
schema.default=noneandschema.constraints=true. - Dgraph: strict mode is a second container started with
--limit mutations=strict.
Found something wrong, or a stricter setting we missed? Open an issue or a PR: the page is updated from this repo.
MIT