GraphForge vs. TerminusDB

How do database branches compare with analyst hypothesis branches?

TerminusDB is a versioned document graph. Each commit appends an immutable delta layer, and the database supports branches, structural diffs, merge, rebase, time travel, and clone, push, and pull between instances. Those are real versioning features, so branching alone does not set GraphForge apart. The difference is what gets versioned. TerminusDB versions data, like Git. GraphForge records investigations: assertions with evidence, competing hypotheses, and algorithms run over alternative states.

Products compared
GraphForgeTerminusDB
ModelEmbedded property graphVersioned document graph
Version0.5.2 stable · 0.6.0-rc.5212.0.8 (2026-10-12)18
LicenseApache-2.0 (since 0.5.0)Apache-2.0 (open source); separate commercial TerminusDB Enterprise
StatusActively maintained2Actively maintained17

Researched 2026-10-19 · 28 primary sources · Head-to-head benchmarks not yet run

On this page

Should I choose TerminusDB or GraphForge?

TerminusDB for Git-like collaboration on documents; GraphForge for graph algorithms and recorded hypotheses.

Choose GraphForge when

  • You need graph algorithms such as PageRank, Louvain, or betweenness, which TerminusDB does not ship.
  • Competing explanations should be explicit, as hypothesis groups with recorded selection events.
  • You want Cypher, Arrow results, and a library inside your code rather than a server.

Choose TerminusDB when

  • Your data is schema-checked JSON documents that several people edit and need Git-like collaboration on.
  • You need push, pull, and clone between database instances.
  • Field-level structural diffs and merges of documents are central to the workflow.

What can each one do?

22 of 26 capabilities differ. For 10, TerminusDB's docs did not confirm support, so they are marked Not established rather than No.

Each capability has a status and a source. “Not established” means we could not confirm the feature from TerminusDB's primary documentation. It does not mean TerminusDB lacks it.

CapabilityGraphForgeTerminusDB
BranchingExperimentalYes
Snapshots / time travelShippedYes
ProvenanceShippedPartial
Competing hypothesesShippedNot established
Graph algorithmsShippedNot established
Capabilities of GraphForge and TerminusDB, with release status and sources
CapabilityGraphForgeTerminusDB
Storage
Property graph modelShipped

Labeled property graph with typed properties.

1
Partial

Linked JSON documents decomposed into RDF triples; not a labeled property-graph model.

19
Document modelNot offered

No document-store model; properties accept nested, heterogeneous values.

1
Yes

JSON and JSON-LD documents via the document API, with schema validation on every write.

1917
Relational tablesNot offered

Not a relational database; results arrive as Arrow tables.

1
Not established

No relational/SQL model documented.

Key-value accessNot offered

No key-value API; data is a property graph.

1
Not established

Not documented.

Columnar storageShipped

Durable projects store graph data as Arrow and Parquet.

4
Not established

Not documented for the core database.

Vector searchShipped

Exact cosine retrieval over embeddings you supply; no ANN index.

5
Extension

VectorLink semantic indexer as separate service; in-process plugin merged to main after 12.0.8.

2728
Full-text searchShipped

Tantivy-backed BM25; hybrid text and vector retrieval through find().

5
Extension

Keyword and hybrid search provided by the LanceDB-based VectorLink indexer, not the core engine.

27
GeospatialNot offered

No spatial types or indexes are documented.

Not established

Not documented.

Execution
Transactional workloads (OLTP)Partial

Local analytical scope; durable atomic writes with queued and optimistic writer modes.

6
Yes

Every commit is ACID with schema validation; readers and writers do not block.

19
Analytical workloads (OLAP)Shipped

Analytical verbs and Cypher over a local columnar project.

7
Not established

Not documented.

Graph traversalShipped

Cypher MATCH patterns plus BFS, DFS, Dijkstra, A*, and reachability.

8
Yes

WOQL datalog with unification and native path() expressions; GraphQL path queries.

2417
Variable-length pathsShipped

Bounded and unbounded patterns and shortestPath().

8
Yes

Path regex operators +, *, {n,m}, |, backward < over document fields.

23
Graph algorithmsShipped

94 native algorithms behind analysis verbs, not Cypher CALL procedures.

37
Not established

No algorithm library documented; own Neo4j comparison points to Neo4j for PageRank/betweenness.

Incremental computationNot offered

Algorithms recompute on request; no incremental mode is documented.

7
Not established

Not documented.

Deployment
Embedded / in-processShipped

In-process library; no server, daemon, or connection string.

9
Not established

Distributed as a server (Docker) and CLI; no in-process library mode documented.

Server modePlanned

Optional remote peer extension planned after 0.6.0.

11
Yes

Docker/docker-compose server on port 6363; also Snap and source installs.

17
WebAssemblyNot planned

A browser WebAssembly engine was declined in ADR 0027.

12
Not established

Not documented.

Distributed clusterNot planned

Distributed clusters and HA replication are out of scope.

9
Paid edition

Clustering listed as a TerminusDB Enterprise feature; open-source model is single-node.

1726
Managed cloudNot planned

No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local.

13
Partial

Gettier managed service in early access; Azure Marketplace listing announced for autumn 2026.

25
Investigation
BranchingExperimental

Research versions, branches, slices, proposals, and forks ship in 0.6.0 release candidates.

14
Yes

Named branches, merge and rebase with git-like semantics.

2021
Snapshots / time travelShipped

Named checkpoints you can diff and revert, plus transaction-time snapshots.

3
Yes

Read any past state by addressing a commit ID instead of a branch.

21
Immutable historyShipped

Knowledge ledger records immutable assertions and evidence lineage.

16
Yes

Each commit appends a content-addressed immutable delta layer; squash can collapse branch history.

1921
ProvenanceShipped

Provenance and knowledge ledger with evidence lineage.

16
Partial

Commit log records identifier, author, timestamp, message; no per-fact lineage model documented.

21
Competing hypothesesShipped

Immutable hypothesis groups with membership and selection events.

3
Not established
Operations
ACID transactionsPartial

Acknowledged-durable atomic publication; docs do not claim general ACID or serializable isolation.

6
Yes

Every commit is atomic, consistent, isolated and durable.

19
ReplicationNot planned

HA replication is out of scope.

9
Partial

Git-style clone, fetch, push, pull between instances; not automatic HA replication.

20

Showing 26 of 26 capabilities.

How is each one built?

TerminusDB: versioned document graph. GraphForge: an embedded property graph library in Rust, persisted as Parquet.

GraphForge

GraphForge is a Rust library that runs inside your process: no daemon, server, or connection string. Projects persist to a local directory as Arrow and Parquet data plus JSON metadata, and every query and analysis verb returns an Apache Arrow table. Algorithms run in the same native engine as Cypher and are called as API verbs (rank, cluster, paths, analyze, similar), not as Cypher procedures.

TerminusDB

TerminusDB is written in Prolog and Rust. Documents are broken into RDF triples held in succinct data structures, and each commit adds a content-addressed, immutable layer, so readers never block writers. Past states are read by addressing a commit ID. Search comes from a separate VectorLink indexer.

TerminusDB tracks versions of data: what changed and who merged it. GraphForge tracks the reasoning: which evidence supports an assertion, which hypothesis was selected, and which checkpoint a result came from. Research versions and branches are experimental in GraphForge 0.6.0 release candidates; checkpoints, provenance, and hypothesis groups are stable.

What the documentation says about TerminusDB

  • Every commit appends a new immutable, content-addressed delta layer recording what was added and removed.19
  • Documents are decomposed into RDF triples stored in succinct data structures, while users work with JSON-LD documents.19
  • Because committed layers are immutable, readers never block writers and writers never block readers.19
  • Clone copies a database's schema, all branches and all layers from one TerminusDB instance to another.20
  • Push succeeds only if the remote branch history is a strict ancestor of the local branch; diverged histories need rebase or merge.20
  • Past states are queried by swapping the branch segment of a resource path for a commit ID, and those reads are read-only.21
  • Structural diffs compare documents, branches or commits at field level rather than as line-based text.22
  • WOQL is a datalog query language that uses Prolog-style unification for variables, producing implicit joins.24
  • The hosted product formerly known as DFRNT Hub and TerminusCMS is now named TerminusDB.25

Which one is faster?

Not measured yet. No head-to-head run has been published, so this page shows no performance chart.

Charts appear only after a reproducible run exists: pinned versions, disclosed hardware, generated data, raw results, and scripts.

No comparative results yet

The suites below apply to GraphForge and TerminusDB. Each will be reported separately per benchmark class, never merged into one leaderboard.

Ingestion and writes

SuiteWorkloadsMetricsStatus
INGESTCSV, JSONL, Parquet, graph edge listsRecords/s, MB/s, total elapsedNot run

Query, analytics, and system

SuiteWorkloadsMetricsStatus
LOOKUPID lookup, indexed property filterp50/p95/p99 latencyNot run
TRAVERSEBFS, k-hop, variable-length pathLatency, visited edges/sNot run
ALGOPageRank, WCC, SCC, Louvain, SSSPTime, memory, correctnessNot run
VERSIONBranch, diff, merge, snapshotLatency, storage overheadNot run
DURABLERestart, recovery, persistenceRecovery time, data integrityNot run

Which graph algorithms ship?

GraphForge: 94 native algorithms. TerminusDB: no named algorithm library in its docs.

Algorithms are listed by name, as documented. Counts of extension modules or third-party libraries are not presented as native coverage.

GraphForge

94 Rust algorithms run in the engine behind five analysis verbs (rank, cluster, paths, analyze, similar) and return Arrow tables.37

A selection of the catalog:

  • PageRank
  • ArticleRank
  • Betweenness (Brandes)
  • Closeness
  • Harmonic closeness
  • Eigenvector
  • HITS
  • CELF influence maximization
  • Triangle count
  • K-core decomposition
  • Louvain
  • Leiden
  • Label propagation
  • Infomap
  • Walktrap
  • Girvan-Newman
  • Weakly connected components
  • Strongly connected components
  • Dijkstra
  • A*
  • Bellman-Ford
  • Yen's k-shortest paths
  • Max flow / min cut
  • Steiner tree
  • Minimum spanning tree
  • Topological sort
  • Node2Vec
  • GraphSAGE
  • Fast random projection
  • Node similarity (Jaccard)

TerminusDB

No graph-algorithm library; graph analysis is expressed as WOQL datalog queries and regular-expression path queries.232426

Delivered as: n/a

No named algorithm catalog was found in the primary documentation.

Where can each one run?

GraphForge runs in-process only. TerminusDB: embedded not established, server yes, cluster paid edition.

ModeGraphForgeTerminusDB
Embedded / in-processShipped

In-process library; no server, daemon, or connection string.

9
Not established

Distributed as a server (Docker) and CLI; no in-process library mode documented.

Server modePlanned

Optional remote peer extension planned after 0.6.0.

11
Yes

Docker/docker-compose server on port 6363; also Snap and source installs.

17
WebAssemblyNot planned

A browser WebAssembly engine was declined in ADR 0027.

12
Not established

Not documented.

Distributed clusterNot planned

Distributed clusters and HA replication are out of scope.

9
Paid edition

Clustering listed as a TerminusDB Enterprise feature; open-source model is single-node.

1726
Managed cloudNot planned

No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local.

13
Partial

Gettier managed service in early access; Azure Marketplace listing announced for autumn 2026.

25
ReplicationNot planned

HA replication is out of scope.

9
Partial

Git-style clone, fetch, push, pull between instances; not automatic HA replication.

20

What will my team write?

GraphForge: openCypher from Python, Node.js, Rust. TerminusDB: WOQL (datalog), GraphQL, Document/REST API.

TerminusDB developers write WOQL, a datalog with Prolog-style unification, or GraphQL, from JavaScript or Python clients over a server API. GraphForge developers write openCypher and call verbs in-process.

Query languages
GraphForge openCypher
TerminusDB WOQL (datalog), GraphQL, Document/REST API
SDKs and bindings
GraphForge Python, Node.js, Rust
TerminusDB JavaScript/TypeScript, Python, Rust (community), Elixir (community)
Written in
GraphForge Rust
TerminusDB Prolog and Rust
Surfaces
GraphForge Python, Node.js, Rust crates, CLI, VS Code extension, Notebooks, Agent skills

What are the trade-offs?

TerminusDB ships no algorithm library and scales on one node. GraphForge's branches are still experimental.

A fair assessment cuts both ways.

GraphForge

  • Branches, slices, proposals, and forks are experimental, available only in 0.6.0 release candidates.
  • No document model or JSON-LD schema checking.
  • No push or pull between instances; projects are shared through the Hub or by copying the directory.

TerminusDB

  • TerminusDB's own Neo4j comparison describes its scalability model as single-node with delta compression.26
  • The same comparison recommends Neo4j when graph algorithms such as PageRank or betweenness centrality are needed.26
  • Squashing a branch combines its commits so intermediate states can no longer be queried on that branch.21
  • Vector and keyword search depend on the separate VectorLink indexer, whose embedded in-process mode is not yet in a tagged release.28
  • Clustering, full JSON-LD/Turtle/RDF-XML document support and faster history queries are reserved for TerminusDB Enterprise.17

How would I migrate?

Load TerminusDB data through Arrow, pandas, or Polars. WOQL queries are rewritten as Cypher.

Into GraphForge

  • GraphForge 0.5.x loads data through Arrow, pandas, and Polars bulk calls or Cypher scripts; it has no native CSV loader.15
  • Algorithms are API verbs (rank, cluster, paths, analyze, similar), not Cypher CALL procedures, so procedure calls become verb calls.7
  • GraphForge passes the full openCypher 2024.3 TCK, so standard openCypher queries are the portable part of a migration.10

Moving from TerminusDB

  • Data enters as schema-validated JSON/JSON-LD documents through the document API or CLI doc insert, after defining classes in the schema graph.17
  • There is no Cypher or SQL; graph queries must be rewritten in WOQL datalog or GraphQL.24
  • Clone, push and pull move full databases with history between TerminusDB instances, but this is TerminusDB-to-TerminusDB only.20

What else do people ask about TerminusDB?

Does TerminusDB support branching and time travel? TerminusDB versions data like Git; GraphForge records hypotheses and evidence; only GraphForge runs graph algorithms.

Does TerminusDB support branching and time travel?

Yes. TerminusDB documents named branches, merge and rebase, structural diffs, and read-only queries against any past commit.

How is GraphForge's hypothesis tracking different from TerminusDB branches?

A TerminusDB branch is an alternative version of the data. GraphForge records competing hypotheses as immutable groups with membership and selection events, linked to assertions and their evidence, so the reasoning is part of the record.

Does TerminusDB run graph algorithms?

It ships no algorithm library. Its own comparison with Neo4j recommends Neo4j when algorithms such as PageRank or betweenness are needed. GraphForge ships 94.

How was this researched?

From 28 primary sources, accessed 2026-10-19, each cited where it is used.

Capabilities were checked against primary documentation, repositories, and release notes, accessed 2026-10-19. Each cell records two things: the capability’s state and the evidence behind it. Every GraphForge cell carries a release status. Shipped means stable 0.5.2; experimental means 0.6.0 release candidates only. GraphForge is pre-1.0 and makes no backward-compatibility guarantee yet.

  • Yes documented natively
  • Partial limited or caveated
  • Extension official extension or module
  • Paid edition paid or enterprise edition only
  • Removed removed, with the version cited
  • Not established not confirmed by our research; not a “no”
  • Shipped in stable 0.5.2
  • Partial shipped with documented limits
  • Experimental in 0.6.0 release candidates only
  • Planned on the roadmap
  • Not offered outside the product's model
  • Not planned explicitly out of scope

Benchmark classes

  1. Equivalent workloads. The same logical operation with the same correctness requirement on every engine.
  2. Native optimized workloads. Each engine's best supported implementation, including projections, indexes, extensions, and native algorithms.
  3. End-to-end application workloads. Time from raw data to a usable result: import, indexing, projection, computation, and output retrieval.

Sources · accessed 2026-10-19

  1. GraphForge repository README — GitHub
  2. GraphForge releases — GitHub
  3. GraphForge 0.5.0 changelog — GitHub
  4. Storage architecture — GraphForge Book
  5. Analytics integration: search and retrieval — GraphForge Guide
  6. ADR 0018: acknowledged durability and isolation — GitHub
  7. Algorithm verbs — GraphForge Book
  8. Cypher guide — GraphForge Guide
  9. openCypher compatibility — GraphForge Reference
  10. openCypher TCK compliance — GraphForge Reference
  11. Remote authority peer extension (#215) — GitHub
  12. ADR 0027: native runtime boundary — GitHub
  13. Release scope and roadmap — GraphForge Docs
  14. Keep research history — GraphForge Guide
  15. Graph construction — GraphForge Guide
  16. Knowledge public API v1 — GraphForge Book
  17. terminusdb/terminusdb README — GitHub
  18. Release v12.0.8 — terminusdb/terminusdb GitHub
  19. What is TerminusDB? — TerminusDB Docs
  20. Git-for-data reference — TerminusDB Docs
  21. Graph database with time travel queries — TerminusDB Docs
  22. Diff and patch operations — TerminusDB Docs
  23. Path Query Reference Guide — TerminusDB Docs
  24. WOQL explanation — TerminusDB Docs
  25. TerminusDB Cloud — TerminusDB Docs
  26. TerminusDB vs Neo4j — TerminusDB Docs
  27. terminusdb-org/vectorlink README — GitHub
  28. Embed vectorlink as a builtin core plugin (PR #2556) — terminusdb/terminusdb GitHub

Try GraphForge on your own graph.

Install the package, open a project, and run the same notebook we measure with.