pre-mvp
Home

Comparison

WWKG vs. the field

How does WWKG compare to established graph databases, the new generation of versioned graph platforms, and decentralized knowledge graph projects? Feature by feature.

For the product overview, visit the official WWKG homepage. WWKG stands for World Wide Knowledge Graph.

The Landscape

Established enterprise triplestores, next-generation versioned databases, P2P graph platforms, decentralized knowledge graph protocols, and data sovereignty platforms.

WWKG

P2P content-addressed encrypted graph database on IPFS

RustTBD

Stardog

Enterprise knowledge graph platform with data virtualization

JavaProprietary

GraphDB

Semantic repository with reasoning and high-availability clustering

JavaProprietary + Free Edition

AllegroGraph

Federated triplestore with neuro-symbolic AI and vector store

LispProprietary + Free Edition

RDFox

High-performance in-memory RDF triplestore with Datalog reasoning

C++Proprietary

Virtuoso

Universal server with SQL/SPARQL hybrid and data virtualization

CProprietary + Open Source Edition

Fluree

Immutable ledger-backed semantic graph database

ClojureOpen Source + Cloud

TerminusDB

Git-for-data graph database with succinct data structures

Prolog / RustApache 2.0

DefraDB

P2P edge-first database on IPLD with MerkleCRDTs

GoOpen Source

Oxigraph

Lightweight single-node SPARQL database in Rust

RustMIT / Apache 2.0

TypeDB

Polymorphic database with custom type system, rewritten in Rust

RustOpen Source + Cloud

ArangoDB

Multi-model database with native graph, document, and key/value support

C++Apache 2.0 + Enterprise

AtomicServer

Event-sourced linked data server with typed schema

RustMIT

OriginTrail

Decentralized knowledge graph on multi-chain blockchain

JS / TSOpen Source + Token

Tentris

Tensor-based triplestore with hypertrie indexes and worst-case optimal joins

C++Apache 2.0 / MIT

Solid

Decentralized personal data pods with linked data and user-controlled access

JS / TSMIT (specs + reference)

GraphQLite

SQLite extension adding graph database capabilities via Cypher

COpen Source (MIT)

LadybugDB

Embedded columnar graph database — 'DuckDB for graphs'

UnknownOpen Source (MIT)

Feature Matrix

Supported Partial Not supported

FeatureWWKGStardogGraphDBAllegroGraphRDFoxVirtuosoFlureeTerminusDBDefraDBOxigraphTypeDBArangoDBAtomicServerOriginTrailTentrisSolidGraphQLiteLadybugDB
Architecture
Peer-to-peer distribution
Content-addressed storage
Offline-capable
No central server required
Conflict-aware distributed convergence
Data Model & Query
RDF / Linked Data native
Subject-clustered pages
Structured + unstructured data
Full-text search
Query Languages
RDF 1.2
SPARQL 1.1
SPARQL 1.2
RDF-star (RDF*)
openCypher
ISO GQL
SQL
Graph Store Protocol
GraphQL
Proprietary query language
Versioning & Branching
Git-style branching
Immutable commits
Time-travel queries
Three-way merge
Delta / diff queries
Security & Encryption
End-to-end encryption
Encryption before storage
Content integrity verification
Decentralized identity (DIDs)
Fine-grained access control (RBAC)
Cryptographic workspace access control
Enterprise & Operations
High-availability clustering
OWL / RDFS reasoning engine
Datalog reasoning
SWRL / RIF rules
LPG reasoning
SHACL / ShEx validation
Data virtualization / federation
Managed cloud offering
Visual graph explorer / GUI tools
Production-ready today
AI & Vector
Built-in vector search
Graph-derived embeddings (KGE)
Vector similarity as reasoning
Versioned vector indexes
Encrypted vector search
Native GraphRAG (single query)
Data Model Breadth
Property graph / LPG support
ACID transactions
Immutable transaction history
Graph algorithms
Indexing & Performance
HAMT Merkle-DAG indexes
Structural sharing between versions
Ad-hoc index creation
Zero-copy block access
In-memory execution

vs. Established Triplestores

Stardog, GraphDB, AllegroGraph, RDFox, Virtuoso

The enterprise incumbents are mature, feature-rich, and battle-tested for centralized deployments. They excel at reasoning (GraphDB, RDFox), data virtualization (Stardog, Virtuoso), and federated querying (AllegroGraph). RDFox — now owned by Samsung — is the performance leader for in-memory Datalog reasoning and incremental materialization, with edge/embedded deployment capabilities. Virtuoso is a hybrid SQL/SPARQL universal server with decades of deployment history and strong data virtualization capabilities.

What they share is a fundamentally centralized architecture: mutable indexes on a single server or cluster. None offer native peer-to-peer distribution, content-addressed storage, end-to-end encryption, or git-style branching. Their recent investments focus on LLM/RAG integration and edge deployment, not architectural innovation.

Their recent vector search additions — Stardog’s similarity models, GraphDB’s embedding plugin, AllegroGraph’s vector store — store externally computed embeddings as node properties. WWKG derives embeddings from the graph itself, versions them alongside the data, encrypts them with the workspace key, and composes vector similarity with ontology reasoning in a single query plan.

WWKG is not competing on enterprise features or ecosystem maturity. It is competing on architecture: the assumption that knowledge graphs should be distributed, versioned, and encrypted by default.

vs. Versioned Graph Databases

Fluree, TerminusDB

Fluree and TerminusDB are the closest conceptual relatives to WWKG. Fluree brings immutability and RDF/SPARQL support with a ledger-backed audit trail. TerminusDB pioneered git-for-data with branches, merges, and time-travel.

Fluree's “decentralized” mode uses blockchain verification rather than true peer-to-peer distribution. TerminusDB has no SPARQL support (using custom WOQL instead) and no native P2P or encryption. Its stewardship was transferred to a three-person company in 2025.

WWKG combines the best of both — immutability and SPARQL from Fluree's world, branching and merging from TerminusDB's — while adding true P2P distribution and end-to-end encryption that neither offers.

vs. P2P Data Platforms

DefraDB, AtomicServer, Ceramic

DefraDB (Source Network) is architecturally the most similar to WWKG — built on IPLD, LibP2P, content-addressed storage, and MerkleCRDTs. But it uses GraphQL, not SPARQL, and targets Web3/dApp developers rather than the semantic web community.

AtomicServer is a fast, Rust-based linked data server with event sourcing, but it is single-node only with no SPARQL and no P2P distribution.

WWKG occupies the unique intersection: IPFS-based P2P distribution with full RDF/SPARQL compatibility. It is the only project that brings the content-addressed, distributed architecture of DefraDB together with the semantic web standards of traditional triplestores.

vs. Decentralized Knowledge Graphs

OriginTrail, The Graph

OriginTrail's Decentralized Knowledge Graph is a production network with enterprise partnerships and blockchain-backed verification. The Graph Protocol indexes blockchain data across a decentralized network of indexers.

Both require blockchain infrastructure and token economics, creating cultural and technical barriers for enterprise and research users who want decentralization without the overhead of Web3 consensus mechanisms and token governance.

WWKG achieves decentralization through cryptography and content-addressing alone — no blockchain, no staking requirement, no consensus protocol. The workspace encryption key is the primary coordination primitive.

vs. Data Sovereignty Platforms

Solid, Inrupt

Solid (Tim Berners-Lee) gives users personal online data stores — “pods” — where they control who can access their linked data. Inrupt provides the commercial Solid platform. Both share WWKG's commitment to user data sovereignty and linked data standards.

Solid pods are server-hosted containers with ACL-based access control. Data is addressed by URL, stored in cleartext on the pod server, and lacks versioning or branching. The pod provider can read everything. There is no content-addressing, no end-to-end encryption, and no offline capability.

WWKG workspaces are the encrypted, content-addressed, version-controlled counterpart to Solid pods. Data is encrypted before it leaves the client, addressed by BLAKE3 hash, and replicated through IPFS without any server needing to read it. Where Solid trusts the pod provider, WWKG trusts only the encryption key holder.

vs. Multi-Model Databases

ArangoDB

ArangoDB combines graph, document, and key/value models in a single engine with its own query language (AQL), ACID transactions, HA clustering, and a managed cloud offering (ArangoGraph). It is a mature, production-ready platform with built-in vector search and a polished web UI.

ArangoDB operates in a fundamentally different world: property graphs with JSON documents, not RDF with semantic web standards. It has no SPARQL support, no content-addressing, no branching or versioning, and no peer-to-peer distribution. Its encryption is server-side at rest, not end-to-end.

ArangoDB’s vector search (ArangoSearch) stores externally computed embeddings alongside documents. WWKG derives embeddings from graph structure, encrypts them end-to-end, and versions them with the commit DAG — capabilities that a document-centric architecture cannot provide.

WWKG and ArangoDB target different audiences. ArangoDB serves teams that want multi-model flexibility in a conventional database. WWKG serves teams that need RDF/SPARQL semantics with distributed, versioned, and encrypted storage that no multi-model database provides.

Clustering & High Availability

Mission-critical deployments without a single point of failure

Traditional graph databases achieve high availability through leader-follower clustering: one node accepts writes, replicas serve reads, and a consensus protocol elects a new leader when one fails. This works, but it requires careful capacity planning, introduces write bottlenecks, and ties availability to a central coordinator.

WWKG takes a fundamentally different approach. Every commit automatically replicates its content-addressed blocks to all connected peers via QUIC. Any node with the blocks can serve reads immediately. In HA mode, writes are synchronously replicated before acknowledgement — the same durability guarantee as synchronous replication in traditional clusters, but without leader election or failover complexity.

Because blocks are immutable and self-verifying (every CID is a BLAKE3 hash of its content), there is no risk of replication corruption. Nodes that go offline catch up automatically through pull-on-reconnect. Conflicts between concurrent writers are surfaced explicitly through the branch model and resolved via three-way merge — no silent last-write-wins, no data loss.

In CAP terms, WWKG favors AP (availability + partition tolerance) with strong eventual consistency through causal commit ordering and deterministic branch merging. This makes it well-suited for geo-distributed deployments, edge nodes, and environments where network partitions are expected rather than exceptional.

Honest Trade-offs

What WWKG deliberately does not do

WWKG is not trying to be everything. It makes deliberate architectural trade-offs in favor of distribution, immutability, and encryption:

  • New entrant. WWKG 1.0 is production-ready but younger than established competitors. The trade-off is a modern architecture without legacy baggage.

The Systems-Language Generation

Oxigraph, TypeDB 3.0, AtomicServer, Tentris, CozoDB

A new generation of graph databases is being written in systems languages: Oxigraph (Rust) for SPARQL compliance, TypeDB 3.0 (Rust) for its polymorphic type system, AtomicServer (Rust) for linked data, Tentris (C++) for tensor-based hypertrie indexing with worst-case optimal joins, CozoDB (Rust) for Datalog. WWKG builds its own SPARQL parser and shares the conviction that systems-level engineering is the right foundation for next-generation database infrastructure. What distinguishes WWKG is the complete architectural rethink — not just a faster engine, but a fundamentally different storage and distribution model.