GraphForge vs. ArangoDB
When should documents and graph relationships share one database?
ArangoDB stores JSON documents, key-value data, and graphs in one engine queried with AQL, and runs as a single server or sharded cluster. Since 3.12, its in-database Pregel algorithms are gone: traversals and path-finding remain, while centrality and community algorithms moved to separate Graph Analytics Engines. The source license also changed to BUSL-1.1. GraphForge is an embedded graph engine with algorithms built in, so the comparison is one shared multimodel database against an analysis-local graph.
| GraphForge | ArangoDB | |
|---|---|---|
| Model | Embedded property graph | Document + graph + key-value |
| Version | 0.5.2 stable · 0.6.0-rc.52 | 3.12.12.1 (2026-09-29)17 |
| License | Apache-2.0 (since 0.5.0) | BUSL-1.1 source from 3.12; Community binaries non-commercial with a 100 GiB cap |
| Status | Actively maintained2 | Actively maintained17 |
On this page
Should I choose ArangoDB or GraphForge?
ArangoDB when documents are the system of record; GraphForge when graph algorithms are the point of the work.
Choose GraphForge when
- You need PageRank, Louvain, or betweenness inside the engine, without a separate analytics product.
- You want an Apache-2.0 library rather than BUSL-licensed binaries with usage caps.
- You want Cypher rather than a proprietary query language.
Choose ArangoDB when
- Your application data is document-shaped, and relationships are one way of querying it.
- You need a sharded cluster with replication and automatic failover.
- AQL joins across documents, key-value data, search views, and graphs in a single query are worth a new query language.
What can each one do?
20 of 26 capabilities differ. For 10, ArangoDB'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 ArangoDB's primary documentation. It does not mean ArangoDB lacks it.
| Capability | GraphForge | ArangoDB |
|---|---|---|
| Storage | ||
| Property graph model | Shipped Labeled property graph with typed properties. 1 | Yes Native graphs over node and edge collections, queried with AQL graph traversals. 2521 |
| Document model | Not offered No document-store model; properties accept nested, heterogeneous values. 1 | Yes Native JSON document store; README calls it multi-model for documents, graphs, key-values. 25 |
| Relational tables | Not offered Not a relational database; results arrive as Arrow tables. 1 | Not established No SQL interface found; queries use AQL over document collections. |
| Key-value access | Not offered No key-value API; data is a property graph. 1 | Yes Key-value access via document _key; README lists key-value as a data model. 25 |
| Columnar storage | Shipped Durable projects store graph data as Arrow and Parquet. 4 | Not established RocksDB storage engine; no columnar storage found in primary docs. |
| Vector search | Shipped Exact cosine retrieval over embeddings you supply; no ANN index. 5 | Yes Faiss-backed vector index since v3.12.4; requires --vector-index startup flag and training on loaded data. 20 |
| Full-text search | Shipped Tantivy-backed BM25; hybrid text and vector retrieval through find(). 5 | Yes ArangoSearch Views and inverted indexes with Analyzers and BM25/TF-IDF ranking. 23 |
| Geospatial | Not offered No spatial types or indexes are documented. | Yes AQL documented as covering geo-spatial data alongside graphs, documents, search, vector. 28 |
| Execution | ||
| Transactional workloads (OLTP) | Partial Local analytical scope; durable atomic writes with queued and optimistic writer modes. 6 | Yes Transactional multi-model server; single-document operations fully ACID including in clusters. 22 |
| Analytical workloads (OLAP) | Shipped Analytical verbs and Cypher over a local columnar project. 7 | Partial Pregel removed in 3.12; graph analytics moved to separate Graph Analytics Engines in the platform. 1824 |
| Graph traversal | Shipped Cypher MATCH patterns plus BFS, DFS, Dijkstra, A*, and reachability. 8 | Yes AQL traversals, shortest path, all shortest paths, k shortest paths, k paths. 21 |
| Variable-length paths | Shipped Bounded and unbounded patterns and shortestPath(). 8 | Yes AQL traversals support configured, optionally variable depth (min..max). 21 |
| Graph algorithms | Shipped 94 native algorithms behind analysis verbs, not Cypher CALL procedures. 37 | Paid edition In-database Pregel algorithms removed in 3.12; Graph Analytics Engines ship with Arango Contextual Data Platform. 1824 |
| Incremental computation | Not offered Algorithms recompute on request; no incremental mode is documented. 7 | Not established No incremental view maintenance or incremental analytics found in primary docs. |
| Deployment | ||
| Embedded / in-process | Shipped In-process library; no server, daemon, or connection string. 9 | Not established Deploy docs list single instance, cluster, OneShard, platform; no embedded library mode. |
| Server mode | Planned Optional remote peer extension planned after 0.6.0. 11 | Yes arangod server with HTTP API; Docker image is the recommended packaging. 2618 |
| WebAssembly | Not planned A browser WebAssembly engine was declined in ADR 0027. 12 | Not established No WebAssembly build found in primary docs. |
| Distributed cluster | Not planned Distributed clusters and HA replication are out of scope. 9 | Yes Sharded cluster deployments with replication and automatic failover; OneShard variant for multi-tenant use. 26 |
| Managed cloud | Not planned No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local. 13 | Yes Arango Managed Platform (AMP), hosted across major cloud providers. 25 |
| Investigation | ||
| Branching | Experimental Research versions, branches, slices, proposals, and forks ship in 0.6.0 release candidates. 14 | Not established No graph/database branching feature found in primary docs. |
| Snapshots / time travel | Shipped Named checkpoints you can diff and revert, plus transaction-time snapshots. 3 | Not established No as-of / time-travel query feature found in primary docs. |
| Immutable history | Shipped Knowledge ledger records immutable assertions and evidence lineage. 16 | Not established No immutable version history feature found in primary docs. |
| Provenance | Shipped Provenance and knowledge ledger with evidence lineage. 16 | Not established No built-in provenance tracking found in primary docs. |
| Competing hypotheses | Shipped Immutable hypothesis groups with membership and selection events. 3 | Not established |
| Operations | ||
| ACID transactions | Partial Acknowledged-durable atomic publication; docs do not claim general ACID or serializable isolation. 6 | Partial Fully ACID on single server/OneShard; cluster gives per-DB-Server atomicity and Local Snapshot Isolation. 22 |
| Replication | Not planned HA replication is out of scope. 9 | Yes Cluster replication with failover; Active Failover and DC2DC removed in 3.12. 2618 |
Showing 26 of 26 capabilities.
How is each one built?
ArangoDB: document + graph + key-value. 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.
ArangoDB
ArangoDB is a C++ server whose documents live in collections; edge collections turn them into graphs. AQL handles traversals, shortest paths, k shortest paths, and k paths natively. ArangoSearch adds inverted-index views with BM25 scoring, and a Faiss-backed vector index arrived in 3.12.4. Graph Analytics Engines run on separate compute, load data into memory, and write results back.
After 3.12, graph analytics in ArangoDB sit outside the database. In GraphForge, traversal, algorithms, and search share one in-process engine and one on-disk project.
What the documentation says about ArangoDB
- Starting with 3.12, ArangoDB source code moved from Apache 2.0 to the Business Source License 1.1.18
- The Pregel distributed graph processing system, its algorithms, PREGEL_RESULT() and the @arangodb/pregel module are unsupported from v3.12 onward.18
- From 3.12.5 the Community Edition includes all former Enterprise Edition features, including SmartGraphs, but commercial use or datasets over 100 GiB still need a license.19
- A Faiss-backed vector index arrived in v3.12.4, supporting cosine and L2, with innerProduct added in v3.12.6.20
- AQL offers five graph query types: traversals, shortest path, all shortest paths, k shortest paths and k paths.21
- ArangoSearch provides Views over inverted indexes with Analyzers and Okapi BM25 / TF-IDF relevance scoring.23
- Graph Analytics Engines run separately from ArangoDB Core on dedicated compute, importing data into memory and writing results back.24
- OneShard deployments are clusters where each database's data is restricted to a single shard, aimed mainly at multi-tenant use.26
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 ArangoDB. Each will be reported separately per benchmark class, never merged into one leaderboard.
Ingestion and writes
| Suite | Workloads | Metrics | Status |
|---|---|---|---|
| INGEST | CSV, JSONL, Parquet, graph edge lists | Records/s, MB/s, total elapsed | Not run |
Query, analytics, and system
| Suite | Workloads | Metrics | Status |
|---|---|---|---|
| LOOKUP | ID lookup, indexed property filter | p50/p95/p99 latency | Not run |
| TRAVERSE | BFS, k-hop, variable-length path | Latency, visited edges/s | Not run |
| JOIN | Multi-relation pattern matching | Latency, rows/s | Not run |
| ALGO | PageRank, WCC, SCC, Louvain, SSSP | Time, memory, correctness | Not run |
| HYBRID | Vector + graph + metadata filters | Recall@k, latency, throughput | Not run |
| SCALE | Increasing graph size and density | Throughput degradation, limits | Not run |
Which graph algorithms ship?
GraphForge: 94 native algorithms. ArangoDB: 11 named in its docs, delivered as separate product.
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)
ArangoDB
AQL provides path-finding natively; in-database Pregel was removed in 3.12, and centrality/community algorithms now run in separate in-memory Graph Analytics Engines included in the Arango Contextual Data Platform.2124
Delivered as: separate product
- Shortest Path
- All Shortest Paths
- k Shortest Paths
- k Paths
- PageRank
- Weakly Connected Components
- Strongly Connected Components
- Betweenness Centrality
- LineRank
- Label Propagation
- Attribute Propagation
Where can each one run?
GraphForge runs in-process only. ArangoDB: embedded not established, server yes, cluster yes.
| Mode | GraphForge | ArangoDB |
|---|---|---|
| Embedded / in-process | Shipped In-process library; no server, daemon, or connection string. 9 | Not established Deploy docs list single instance, cluster, OneShard, platform; no embedded library mode. |
| Server mode | Planned Optional remote peer extension planned after 0.6.0. 11 | Yes arangod server with HTTP API; Docker image is the recommended packaging. 2618 |
| WebAssembly | Not planned A browser WebAssembly engine was declined in ADR 0027. 12 | Not established No WebAssembly build found in primary docs. |
| Distributed cluster | Not planned Distributed clusters and HA replication are out of scope. 9 | Yes Sharded cluster deployments with replication and automatic failover; OneShard variant for multi-tenant use. 26 |
| Managed cloud | Not planned No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local. 13 | Yes Arango Managed Platform (AMP), hosted across major cloud providers. 25 |
| Replication | Not planned HA replication is out of scope. 9 | Yes Cluster replication with failover; Active Failover and DC2DC removed in 3.12. 2618 |
What will my team write?
GraphForge: openCypher from Python, Node.js, Rust. ArangoDB: AQL.
ArangoDB uses AQL, its own declarative language, with drivers for Java, Go, JavaScript, and Python. GraphForge uses openCypher, with Python, Node.js, and Rust bindings that return Arrow tables, so porting in either direction means rewriting queries.
- Query languages
- GraphForge openCypher
- ArangoDB AQL
- SDKs and bindings
- GraphForge Python, Node.js, Rust
- ArangoDB Java, Go, JavaScript (arangojs), Python (python-arango)
- Written in
- GraphForge Rust
- ArangoDB C++
- Surfaces
- GraphForge Python, Node.js, Rust crates, CLI, VS Code extension, Notebooks, Agent skills
What are the trade-offs?
ArangoDB moved algorithms out of the core in 3.12 and licenses under BUSL. GraphForge has no document model or cluster.
A fair assessment cuts both ways.
GraphForge
- No document model; nested values are graph properties, not a document store.
- No sharded cluster or replication.
- No native CSV or JSON importer comparable to arangoimport; data arrives through dataframes or Arrow.
ArangoDB
- 3.12 removed native Windows and macOS packages, Active Failover, DC2DC replication, LDAP auth and the VelocyStream protocol.18
- In a cluster, distributed-commit atomicity is not guaranteed and isolation is only Local Snapshot Isolation per DB-Server.22
- Stream/JavaScript transactions are capped at 128 MB per DB-Server with a maximum 120-second idle timeout.22
- Vector indexes need training on existing data; vectors added later are assigned to existing centroids and search quality can degrade.20
- Prepackaged Community binaries forbid commercial use and cap production datasets at 100 GiB.18
How would I migrate?
Load ArangoDB data through Arrow, pandas, or Polars. AQL 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
What else do people ask about ArangoDB?
Does ArangoDB still support Pregel? Pregel is gone since 3.12, the source is BUSL-1.1, and GraphForge replaces only the graph-analysis part.
Does ArangoDB still support Pregel?
No. ArangoDB's 3.12 release notes state that Pregel, its algorithms, PREGEL_RESULT(), and the @arangodb/pregel module are no longer supported. Path-finding stays in AQL; other algorithms moved to Graph Analytics Engines.
Is ArangoDB still open source?
From 3.12, the source code is licensed under BUSL-1.1, and prepackaged Community binaries forbid commercial use and cap production datasets at 100 GiB. GraphForge is Apache-2.0.
Can GraphForge replace ArangoDB?
Only for graph analysis. GraphForge does not store documents or run as a cluster. Teams often keep application data in a database like ArangoDB and export the graph they want to analyze.
How do I move an ArangoDB graph into GraphForge?
Export vertex and edge collections as JSON or CSV, load them into dataframes, and publish them into a GraphForge project. AQL queries need to be rewritten in Cypher.
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
- Equivalent workloads. The same logical operation with the same correctness requirement on every engine.
- Native optimized workloads. Each engine's best supported implementation, including projections, indexes, extensions, and native algorithms.
- 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
- GraphForge repository README — GitHub
- GraphForge releases — GitHub
- GraphForge 0.5.0 changelog — GitHub
- Storage architecture — GraphForge Book
- Analytics integration: search and retrieval — GraphForge Guide
- ADR 0018: acknowledged durability and isolation — GitHub
- Algorithm verbs — GraphForge Book
- Cypher guide — GraphForge Guide
- openCypher compatibility — GraphForge Reference
- openCypher TCK compliance — GraphForge Reference
- Remote authority peer extension (#215) — GitHub
- ADR 0027: native runtime boundary — GitHub
- Release scope and roadmap — GraphForge Docs
- Keep research history — GraphForge Guide
- Graph construction — GraphForge Guide
- Knowledge public API v1 — GraphForge Book
- CHANGELOG (v3.12.12.1) — arangodb/arangodb on GitHub
- Incompatible changes in ArangoDB 3.12 — Arango docs
- Features and Improvements in ArangoDB 3.12 — Arango docs
- Vector indexes — Arango docs
- Graph queries — Arango docs
- Transaction limitations — Arango docs
- ArangoSearch — Arango docs
- Graph Analytics — Arango docs
- arangodb/arangodb README — GitHub
- Deploy ArangoDB — Arango docs
- arangoimport — Arango docs
- ArangoDB 3.12 documentation home — Arango docs
Try GraphForge on your own graph.
Install the package, open a project, and run the same notebook we measure with.