GraphForge vs. CozoDB
How do Datalog and graph algorithms compare with Cypher queries and analysis verbs?
CozoDB is a Rust, embeddable relational-graph-vector database queried with CozoScript, a Datalog. It has built-in algorithms (PageRank, Louvain, A*, betweenness), HNSW vectors, opt-in time travel, and pluggable storage backends. Its last release was v0.7.6 in December 2023, and its maintenance status is unclear. GraphForge offers a similar embedded, algorithm-rich design with Cypher and an actively released engine.
On this page
Should I choose CozoDB or GraphForge?
CozoDB when recursive Datalog fits the problem best; GraphForge for Cypher on an actively released engine.
Choose GraphForge when
- You want a project with stable 2026 releases and a published openCypher TCK result.
- Your team already knows Cypher, or will hand queries to tools and agents that do.
- Your investigation needs provenance and recorded competing hypotheses, not only time travel.
Choose CozoDB when
- Recursive logic is the heart of the problem and Datalog expresses it most naturally.
- You need an embedded engine on Android, iOS, Java, Clojure, or Go.
- You want HNSW proximity graphs exposed as graphs that algorithms can run on.
What can each one do?
18 of 26 capabilities differ. For 8, CozoDB'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 CozoDB's primary documentation. It does not mean CozoDB lacks it.
| Capability | GraphForge | CozoDB |
|---|---|---|
| Storage | ||
| Property graph model | Shipped Labeled property graph with typed properties. 1 | No Relational model by design; README explicitly rejects the labelled-property graph model and stores graphs as relations. 17 |
| Document model | Not offered No document-store model; properties accept nested, heterogeneous values. 1 | Partial JSON values supported since v0.7; no document-store model. 17 |
| Relational tables | Not offered Not a relational database; results arrive as Arrow tables. 1 | Yes Stored relations queried with Datalog; ad-hoc joins. 1720 |
| Key-value access | Not offered No key-value API; data is a property graph. 1 | Not established Backends are key-value stores internally, but no KV API is exposed to users. |
| Columnar storage | Shipped Durable projects store graph data as Arrow and Parquet. 4 | No Storage layer defines a row-oriented binary data format. 17 |
| Vector search | Shipped Exact cosine retrieval over embeddings you supply; no ANN index. 5 | Yes HNSW indexes on vector columns (L2, Cosine, IP) usable inside Datalog queries. 22 |
| Full-text search | Shipped Tantivy-backed BM25; hybrid text and vector retrieval through find(). 5 | Yes FTS index added in v0.7; MinHash-LSH index for near-duplicate search. 22 |
| Geospatial | Not offered No spatial types or indexes are documented. | Not established No geospatial types or indexes found in docs. |
| Execution | ||
| Transactional workloads (OLTP) | Partial Local analytical scope; durable atomic writes with queued and optimistic writer modes. 6 | Yes Transactional mutations with MVCC and multi-statement transactions. 1720 |
| Analytical workloads (OLAP) | Shipped Analytical verbs and Cypher over a local columnar project. 7 | Partial Scans and aggregations over row-oriented relations; no columnar or vectorized engine. 17 |
| Graph traversal | Shipped Cypher MATCH patterns plus BFS, DFS, Dijkstra, A*, and reachability. 8 | Yes Recursive Datalog rules, including recursion through safe aggregations. 17 |
| Variable-length paths | Shipped Bounded and unbounded patterns and shortestPath(). 8 | Yes Via recursive rules and fixed rules such as ShortestPathBFS and KShortestPathYen; no path-pattern syntax. 1719 |
| Graph algorithms | Shipped 94 native algorithms behind analysis verbs, not Cypher CALL procedures. 37 | Yes Built-in fixed rules (PageRank, Louvain, etc.) behind the graph-algo feature; absent from WASM builds. 19 |
| Incremental computation | Not offered Algorithms recompute on request; no incremental mode is documented. 7 | Not established No incremental view maintenance documented. |
| Deployment | ||
| Embedded / in-process | Shipped In-process library; no server, daemon, or connection string. 9 | Yes Runs in-process from many host languages, including mobile. 17 |
| Server mode | Planned Optional remote peer extension planned after 0.6.0. 11 | Yes Standalone HTTP server executable (cozo-bin) for client-server mode. 17 |
| WebAssembly | Not planned A browser WebAssembly engine was declined in ADR 0027. 12 | Partial Browser WASM build supports only the in-memory backend and omits graph algorithms. 1719 |
| Distributed cluster | Not planned Distributed clusters and HA replication are out of scope. 9 | Partial TiKV backend gives distributed storage; available only from Rust source and the standalone server. 17 |
| Managed cloud | Not planned No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local. 13 | Not established No managed service found. |
| Investigation | ||
| Branching | Experimental Research versions, branches, slices, proposals, and forks ship in 0.6.0 release candidates. 14 | Not established No branching feature documented. |
| Snapshots / time travel | Shipped Named checkpoints you can diff and revert, plus transaction-time snapshots. 3 | Yes Opt-in per relation via a Validity key column; queries take an @ timestamp. 21 |
| Immutable history | Shipped Knowledge ledger records immutable assertions and evidence lineage. 16 | Partial Validity relations keep assertion/retraction history; non-validity relations are updated in place. 21 |
| Provenance | Shipped Provenance and knowledge ledger with evidence lineage. 16 | Not established No provenance feature documented. |
| 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 Atomic snapshot-isolated transactions with commit-time conflict checks; durability depends on backend. 20 |
| Replication | Not planned HA replication is out of scope. 9 | Not established No replication feature documented beyond the TiKV storage option. |
Showing 26 of 26 capabilities.
How is each one built?
CozoDB: relational-graph-vector (datalog). 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.
CozoDB
CozoDB stores relations in in-memory, SQLite, RocksDB, Sled, or TiKV backends. Each script runs in its own snapshot transaction. Graph algorithms are fixed rules compiled in with the graph-algo feature, which every prebuilt binary except WASM includes. Time travel is enabled per relation by a Validity key column.
Both are embedded engines with built-in algorithms. CozoDB treats the graph as one shape a relation can take and queries it with logic. GraphForge treats the property graph as the model and queries it with patterns.
What the documentation says about CozoDB
- CozoDB describes itself as a transactional relational-graph-vector database that uses Datalog for queries.17
- Cozo supports in-memory, SQLite, RocksDB, Sled and TiKV storage backends.17
- Cozo calls itself embeddable rather than embedded, since it can also run in client-server mode.17
- Built-in fixed rules include PageRank, Louvain community detection, Yen's k-shortest paths, A* and betweenness centrality.19
- A relation becomes eligible for time travel when the last part of its key has the Validity type, a timestamp plus an assertion flag.21
- HNSW indexes support L2, Cosine and inner-product distance and can be used inside normal Datalog queries.22
- Each script runs in its own transaction, using a snapshot, and commits only if there are no conflicts.20
- The README says HNSW proximity graphs are exposed as regular graphs, so community detection and other algorithms can run on them.17
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 CozoDB. 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 |
| ALGO | PageRank, WCC, SCC, Louvain, SSSP | Time, memory, correctness | Not run |
| HYBRID | Vector + graph + metadata filters | Recall@k, latency, throughput | Not run |
| EMBED | Startup, load, reopen, binary size | ms, MiB, disk footprint | Not run |
Which graph algorithms ship?
GraphForge: 94 native algorithms. CozoDB: 19 named in its docs, delivered as bundled.
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)
CozoDB
Graph algorithms ship as built-in CozoScript fixed rules, compiled in with the graph-algo feature flag that all prebuilt binaries except WASM include.19
Delivered as: bundled
- PageRank
- ConnectedComponents
- StronglyConnectedComponent
- MinimumSpanningForestKruskal
- MinimumSpanningTreePrim
- TopSort
- ShortestPathBFS
- ShortestPathDijkstra
- KShortestPathYen
- ShortestPathAStar
- BreadthFirstSearch
- DepthFirstSearch
- ClusteringCoefficients
- CommunityDetectionLouvain
- LabelPropagation
- DegreeCentrality
- ClosenessCentrality
- BetweennessCentrality
- RandomWalk
Where can each one run?
GraphForge runs in-process only. CozoDB: embedded yes, server yes, cluster partial.
| Mode | GraphForge | CozoDB |
|---|---|---|
| Embedded / in-process | Shipped In-process library; no server, daemon, or connection string. 9 | Yes Runs in-process from many host languages, including mobile. 17 |
| Server mode | Planned Optional remote peer extension planned after 0.6.0. 11 | Yes Standalone HTTP server executable (cozo-bin) for client-server mode. 17 |
| WebAssembly | Not planned A browser WebAssembly engine was declined in ADR 0027. 12 | Partial Browser WASM build supports only the in-memory backend and omits graph algorithms. 1719 |
| Distributed cluster | Not planned Distributed clusters and HA replication are out of scope. 9 | Partial TiKV backend gives distributed storage; available only from Rust source and the standalone server. 17 |
| Managed cloud | Not planned No hosted compute. The graphforge.sh Hub distributes repositories; execution stays local. 13 | Not established No managed service found. |
| Replication | Not planned HA replication is out of scope. 9 | Not established No replication feature documented beyond the TiKV storage option. |
What will my team write?
GraphForge: openCypher from Python, Node.js, Rust. CozoDB: CozoScript (Datalog dialect).
CozoScript is expressive but unfamiliar to most teams, and pre-1.0 Cozo promises no syntax or storage stability. GraphForge is pre-1.0 too, but uses the widely known openCypher language.
- Query languages
- GraphForge openCypher
- CozoDB CozoScript (Datalog dialect)
- SDKs and bindings
- GraphForge Python, Node.js, Rust
- CozoDB Rust, Python, Node.js, Browser (WebAssembly), Java, Clojure, Android, Swift (iOS/macOS), Go, C FFI, Standalone HTTP server
- Written in
- GraphForge Rust
- CozoDB Rust
- Surfaces
- GraphForge Python, Node.js, Rust crates, CLI, VS Code extension, Notebooks, Agent skills
What are the trade-offs?
CozoDB has had no release since December 2023. GraphForge has no Datalog surface or mobile bindings.
A fair assessment cuts both ways.
GraphForge
- No Datalog or logic-programming query surface.
- No mobile bindings (Android, iOS) or WebAssembly build.
- Pre-1.0, with no backward-compatibility guarantee yet.
CozoDB
- The README states versions before 1.0 do not promise syntax/API stability or storage compatibility; the latest release is 0.7.6.17
- The project has had no release since December 2023, and a December 2025 issue asking whether it is maintained has no maintainer response.24
- The browser WASM build is in-memory only and is compiled without the graph-algo feature.19
- Data files for Cozo's SQLite backend use Cozo's own binary row format and cannot be queried with SQL.17
How would I migrate?
Load CozoDB data through Arrow, pandas, or Polars. CozoScript 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 CozoDB
- Datalog rules in CozoScript have no direct Cypher equivalent; recursive rules must be rewritten as MATCH path patterns or procedure calls.17
- export_relations returns a consistent snapshot of named relations as a dict of rows, and backup() writes a standalone SQLite-based Cozo database.25
- CsvReader and JsonReader fixed rules load CSV/JSON files into queries, which helps round-trip node and edge relations through flat files.19
What else do people ask about CozoDB?
Is CozoDB still maintained? CozoDB's maintenance is unclear; its algorithms are absent from the WASM build; most queries translate to Cypher.
Is CozoDB still maintained?
It is unclear. The last release is v0.7.6 from December 2023, and a December 2025 issue asking whether the project is maintained had no maintainer reply when we checked.
Does CozoDB have graph algorithms?
Yes. Fixed rules include PageRank, connected components, Louvain, Yen's k-shortest paths, A*, and betweenness, but they are absent from the WebAssembly build.
Can I translate CozoScript to Cypher?
Pattern-style queries usually translate. Recursive rules map to variable-length paths or to GraphForge path verbs, and fixed-rule algorithm calls map to verb calls.
How was this researched?
From 25 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
- cozodb/cozo README — GitHub
- Releases · cozodb/cozo — GitHub
- Utilities and algorithms — CozoDB v0.7 docs
- Stored relations and transactions — CozoDB v0.7 docs
- Time travel — CozoDB v0.7 docs
- Proximity searches (HNSW, MinHash-LSH, FTS) — CozoDB v0.7 docs
- Commits · cozodb/cozo (main) — GitHub
- Is cozo still being maintained? · Issue #301 · cozodb/cozo — GitHub
- Beyond CozoScript (export/import/backup APIs) — CozoDB v0.7 docs
Try GraphForge on your own graph.
Install the package, open a project, and run the same notebook we measure with.