ArcFlow
Company
Managed Services
Markets
  • News
  • LOG IN
  • GET STARTED

OZ brings Visual Intelligence to physical venues, a managed edge layer that lets real-world environments see, understand, and act in real time.

Talk to us

ArcFlow

  • World Models
  • Sensors

Managed Services

  • OZ VI Venue 1
  • Case Studies

Markets

  • Sports
  • Broadcasting
  • Robotics

Company

  • About
  • Technology
  • Careers
  • Contact

Ready to see it live?

Talk to the OZ team about deploying at your venues, from a single pilot match to a full regional rollout.

Schedule a deployment review

© 2026 OZ. All rights reserved.

LinkedIn
ArcFlow Docs
Start
  • Quickstart
  • Installation
  • Bindings
  • Platforms
  • Get Started
  • Cookbook
Concepts
  • World Model
  • Graph Model
  • Evidence Model
  • Observations
  • Confidence & Provenance
  • Proof Artifacts & Gates
  • SQL vs GQL
  • Graph Patterns
  • Parameters
  • Query Results
  • Persistence & WAL
  • Snapshot-Pinned Reads
  • Error Handling
  • Execution Models
  • Causal Edges
  • Adapter Discipline
  • Time Decay
  • Layers
  • 1. World Store
  • 1a. World Store · Smart Reader
  • 2. Perception Lake
  • 3. World Graph
  • 4. Query Engine
  • 5. Live Surface
  • 6. Event Bus
  • 7. Behavior Engine
  • 8. Algorithm Library
  • Virtual Computed Columns
  • Threading Model
  • Typed ID Contract
  • The Information Layer
  • The Memory Engine
WorldCypher
  • Overview
  • Execution Options
  • Statements
  • MATCH
  • WHERE
  • RETURN
  • OPTIONAL MATCH
  • CREATE
  • SET
  • MERGE
  • DELETE
  • REMOVE
  • ASOF JOIN
  • CREATE NODE LABEL
  • CREATE PROGRAM
  • CREATE TRIGGER
  • CREATE LIVE VIEW
  • CREATE WINDOW
  • CREATE DECAY POLICY
  • REFINE EDGE / REPROCESS EDGES
  • Sessions & Transactions
  • Composition
  • WITH
  • UNION
  • UNWIND
  • CASE
  • FOREACH
  • Schema
  • Schema Overview
  • Indexes
  • Constraints
  • Functions
  • Built-in Functions
  • Aggregations
  • Procedures
  • Shortest Path
  • EXPLAIN
  • PROFILE
  • Temporal Queriesfacet
  • Spatial Queriesfacet
  • Algorithmsfacet
  • Triggers
Capabilities
  • Live Queries
  • Vector Search
  • Trusted RAG
  • Spatial Knowledge
  • Temporal
  • Behavior Graphs
  • Graph Algorithms
  • Skills
  • CREATE SKILL
  • PROCESS NODE
  • REPROCESS EDGES
  • Sync
  • Programs
  • GPU Acceleration
  • Agent-Native
  • MCP Server
  • Event Sourcing
  • Intent Relay
  • Event Bus
Use Cases
  • Agent Tooling
  • Trusted RAG
  • Knowledge Management
  • Behavior Graphs
  • Autonomous Systems
  • Physical AI
  • Digital Twins
  • Robotics & Perception
  • Sports Analytics
  • Grounded Neural Objects
  • Fraud Detection
Walkthroughs
    Guides
  • Agent Integration
  • Building a World Model
  • Modeling a Social Graph
  • Build a RAG Pipeline
  • Using Skills
  • Behavior Graphs
  • Swarm & Multi-Agent
  • Fleet Coordination
  • From SQL to GQL
  • Filesystem Workspace
  • Data Quality
  • Code Intelligence
  • Scale Patterns
  • Lakehouse Fast-Path
  • Python Executor SDK
  • Tutorials
  • Knowledge Graph
  • Entity Linking
  • Vector Search
  • Graph Algorithms
  • Recipes
  • CRUD
  • Multi-MATCH
  • MERGE (Upsert)
  • Full-Text Search
  • Batch Projection
  • Multi-Source Observation
  • Sports Analytics
Operations
  • CLI
  • REPL Commands
  • Snapshot & Restore
  • Filesystem Projection
  • Plugin Management
  • Agent Governance
  • Server Modes & PG Wire
  • Persistence (ops)
  • Import & Export
  • Deployment
  • Deployment Modes
  • Daemon (UDS)
  • Why not Docker
  • Architecture
  • Engine Architecture
  • Cloud Architecture
  • Sync Protocol (Deep Dive)
  • World Graph Substrate (Preview)
Reference
  • TypeScript API
  • Glossary
  • Naming & Domain Map
  • Data Types
  • Operators
  • Error Codes
  • GQL Reference
  • Known Issues
  • Versioning
  • Licensing
  • Conformance
  • GQL Conformance
  • openCypher TCK
GQL Reference
    Conformance
  • Conformance Dashboard
  • openCypher TCK Results
  • Features
  • MATCH Basic
  • CREATE Nodes Edges
  • SET REMOVE Properties
  • DELETE Detach DELETE
  • RETURN WITH WHERE
  • Order BY Limit Skip
  • Order BY Nulls First Last
  • UNWIND
  • Aggregate Functions
  • OPTIONAL MATCH
  • Variable Length Paths
  • Label OR AND NOT Expressions
  • Label Wildcard
  • Quantified Path Sugar
  • Path Modes Walk Trail Simple Acyclic
  • Shortest Path Variants
  • IS Labeled Predicate
  • Element ID Function
  • IS Type Predicate
  • Binary Literals
  • Line Comments Solidus
  • Line Comments Minus
  • GQLSTATUS Result Codes
  • GQL Error Code Mapping
  • Transaction Control Syntax
  • SET Session
  • Conditional Execution WHEN THEN ELSE
  • RETURN NEXT Pipeline
  • Primary Key Constraint
  • Unique Constraint
  • Deterministic MERGE Via PK
  • Undirected Edge MATCH
  • Cast Type Conversion
  • GQL Directories
  • Multiple Labels Per Node
  • GQL Flagger
  • NEXT Linear Composition
  • Cardinality Function
  • INT64 BIGINT Type Names
  • FLOAT64 Double Type Names
  • Log10 Log2 Functions
  • Trim Leading Trailing Both
  • FILTER Clause
  • LET Statement
  • Group BY Explicit
  • EXCEPT SET Operations
  • INTERSECT SET Operations
  • ALL Different Predicate
  • Same Predicate
  • Property Exists Function
  • Path Variable Binding
  • USE Graph Clause
  • FOR IN List
  • Typed Temporal Literals
  • Session SET Value Params
  • Typed List Annotations
  • arcflow.cosine() function
  • arcflow.embed() function
  • arcflow.similar() procedure
  • arcflow.graphrag() procedure
  • ArcFlow Extensions
  • LIVE Queries
  • Reactive Write-Back Views
  • Evidence Algebra
  • Relationship Skills
  • AI Function Namespace
  • Graph Embedding Algorithms
  • ASOF JOIN
  • Durable Workflows
  • Incremental Z-Set Engine
  • GPU GraphBLAS
  • Triggers
  • HNSW Vector Index
  • Extensions Moat

Naming & Domain Map

ArcFlow is a world-model engine — one in-process Rust binary that holds the durable object substrate, the observation landing zone, the typed entity layer, the query engine, the live surface, the event bus, the behavior engine, and the algorithm library. The eight layers each have an industry analog; the names below are the canonical labels for each one. Use these consistently in docs, marketing, API names, log messages, and commit subjects.

The World Model story (one sentence)#

ArcFlow is the world-model engine: a single in-process store that remembers what was observed, infers what is happening, predicts what comes next, and lets agents query, subscribe, fuse, and act — all in one address space.

Everything below is a way of saying what part of this story you mean.

The brand stack#

The names below have a hierarchy. ArcFlow is the hero; everything else is positioning context around the hero.

OZ                                  ← destination brand (the company)
 │                                    "Where worlds are built."
 │
 ├── ArcFlow                ★ HERO  ← the engine product (the flagship)
 │    │                               "The blazing-fast graph engine
 │    │                                for modeling the real world."
 │    │                               In-process. Single binary. Runs on
 │    │                               a laptop. Free forever, locally.
 │    │                               Loved by agents.
 │    │
 │    │  ─── eight architectural layers ───
 │    │
 │    ├── World Store        (L1)   ← storage substrate (internal)
 │    ├── Perception Lake    (L2)   ← observation landing (reserved)
 │    ├── World Graph        (L3) ★ ← typed entity layer — carries hero
 │    ├── Query Engine       (L4)
 │    ├── Live Surface       (L5)
 │    ├── Event Bus          (L6)
 │    ├── Behavior Engine    (L7)
 │    └── Algorithm Library  (L8)
 │
 └── OZ Cloud              ◇ SIDE  ← managed deployment of the engine
                             STORY   "The same ArcFlow engine, hosted."
                                     Backups, encryption, audit log,
                                     region selection, zero ops.
                                     Demonstrates OZ's service competence;
                                     never the hero pitch.

Reading rules for anyone writing about ArcFlow:

  • ArcFlow leads. Every external surface (docs, README, marketing, conference talks, CLI help, error messages) leads with the engine.
  • World Store is internal substrate. It has a brand-clean name so the engine's module tree is navigable, but it is not a sellable SKU and it is never the hero. Phrases like "the World Store product" or "consume the World Store standalone" are anti-pattern.
  • World Graph is the hero layer. When a section needs to point at "what ArcFlow is modeling," it points here.
  • OZ Cloud is a side story. It is mentioned where deployment context warrants; it is never headlined as the pitch. The cloud is the convenience tier; the engine is the product.
  • arcflow-mount is proof, not pitch. When described, it is evidence that the engine is loved by agents (bash + find + cat work natively over typed memory) — not a separate brand surface.

Domain map#

#LayerIndustry analogsWhat it is in ArcFlowCanonical nameCypher / API surface
1Storage substrate (internal)(no consumer-facing analog — internal substrate, not a sellable surface)Iceberg-shaped manifests + WAL segments + snapshots + content-addressed parquet blocks; addressed by lake:// URIsWorld Storelake://, manifest commit, snapshot pin
2Observation landingBronze tier in medallion architecture; raw-events bucketAppend-only observation rows with observation-time stamps; the immutable "raw input" planePerception Lakevirtual-label scans, observation-time queries
3Typed entity layer (★ hero layer)Neo4j, Memgraph, KuzuDBProperty-graph nodes + edges + indexes; mission tiers (observed / inferred / predicted); HLC provenance; the catalog that binds graph schema to the substrateWorld GraphMATCH, CREATE, MERGE, indexes, constraints
4Query engineMemgraph engine, DuckDB engine, VeloxCypher / ISO GQL compiler + planner + executorQuery EngineEXPLAIN, PROFILE, prepared statements
5Live surfaceMaterialize, RisingWave, Flink standing queriesZ-set-incremental standing queries + materialised viewsLive SurfaceCREATE LIVE VIEW, LIVE CALL, LIVE MATCH, db.subscribe()
6Event busNATS, Inngest (events), Redis StreamsTopics + durable consumers + wildcard subscriptions + sync request-replyEvent BusCREATE TOPIC, PUBLISH, SUBSCRIBE, SUBSCRIBE PATTERN, bus.request()
7Behavior engineUnreal Engine 5 behavior trees, Unity ML-Agents, TemporalBehavior-tree ticks, durable workflows, triggers, programsBehavior EngineCREATE TRIGGER, CREATE PROGRAM, CALL behavior.tick()
8Algorithm libraryNeo4j GDS, cuGraph, NetworkX29+ graph + sensor-fusion + spatial algorithmsAlgorithm LibraryCALL algo.*, CALL algo.fusion.*, CALL algo.spatial.*

Cross-cutting surfaces (not architectural layers):

Cross-cutting surfaces are not numbered layers. They span multiple layers and are governed by their own conventions — PAT-0014 for bindings, the Durability Lane doctrine for fsync ordering.

SurfaceWhat it isCanonical name
WAL + snapshot replayAsync WAL + checkpointing + hot backup; the durability protocol that spans World Store and World GraphDurability
Language wrappersC ABI + Python (ctypes), Node (napi-rs), Rust SDKBindings

The umbrella label for the whole binary is the world-model engine or the ArcFlow engine — never just "the graph database" (which under-sells the live + event-bus + behavior + algorithm layers) and never just "the event broker" (which under-sells the graph + query

  • behavior layers).

How the layers compose#

                ┌──────────────────────────────────────────────────┐
                │                                                  │
                │                ArcFlow Engine                    │
                │         (one in-process Rust binary)             │
                │                                                  │
   CALL algo.* ►│  8. Algorithm Library  ──► reads/writes          │
                │              │                                   │
                │              ▼                                   │
   Tick      ──►│  7. Behavior Engine    ──► graph mutations       │
                │              │                                   │
                │              ▼                                   │
   Publish   ──►│  6. Event Bus          ──► topics + DLQ          │
                │              │                                   │
                │              ▼                                   │
   Subscribe ──►│  5. Live Surface       ──► standing-query deltas │
                │              │                                   │
                │              ▼                                   │
   Cypher /  ──►│  4. Query Engine                                 │
   GQL          │              │                                   │
                │              ▼                                   │
                │  3. World Graph        ◄── typed entities        │
                │              │             (view over Store)     │
                │              ▼                                   │
                │  2. Perception Lake    ◄── observation landing   │
                │              │             (Bronze tier)         │
                │              ▼                                   │
   lake://   ──►│  1. World Store        ◄── object substrate      │
                │                            (manifests, WAL,      │
                │                            snapshots, parquet)   │
                │                                                  │
                └──────────────────────────────────────────────────┘
                     ▲
                     │ Bindings (Python / TS / C / Rust)
                     │
                  Application

Every layer reads and writes the same GraphStore. No serialisation between modules. No second store to operate.

URI schemes#

Three schemes, three roles, no overlap. Each scheme is parsed by a dedicated typed parser; cross-scheme resolution is explicit via the catalog.

SchemeRoleScopeStatus
oz://Brand-level URI scheme. Typed entity handles across all OZ products. Today addresses ArcFlow's workspace / snapshot / label / edge / catalog / partition.Brand surface (Layer 3+; extensible across future OZ products)stable
lake://Storage-substrate URI scheme. Layer 1 only. Bytes, parquet, manifests, glob templates. Resolves to s3:// / file:// / gs:// underneath.World Store (Layer 1)stable
world://Individual typed entity URI scheme. Stable-ID addressing for single nodes or edges with mission tier + provenance.World Graph (Layer 3), entity-levelreserved

oz://lake/<path> is rejected — the oz:// parser does not accept lake as an authority. lake://workspace/<id>/... is rejected — workspaces are typed handles, not storage paths.

Naming-coherence rules#

Six rules that bind the namespace decisions to code review:

  1. Layer name = customer-facing brand of that layer. "World Store", "World Graph". Two words, title-case, no jargon.
  2. Module path = lowercase, underscore-joined, single token where possible. worldstore::*, worldgraph::*. The module names are the lowercase form of the layer names — one logical concept per identifier.
  3. URI scheme = lowercase, brand-scoped or substrate-scoped, no product names. oz:// for brand, lake:// for substrate. Never arcflow:// (conflates engine identity with substrate identity).
  4. Cross-product URIs go under oz://. Cross-substrate URIs never. oz://studio/... is forward-compatible if OZ ever ships a Studio surface. lake://studio/... is not — lake:// doesn't know what a Studio is.
  5. Internal Rust module names never carry the product name redundantly. Internally, modules are worldstore::*, not arcflow_worldstore::*. The crate prefix carries product identity; modules carry layer identity.
  6. Brand surfaces use full names; module paths use short names. Docs and marketing say "World Store"; code says worldstore. Two-word brand renders as one-word module to keep paths short and the brand distinct from the path.

Banned words#

These terms are not used in user-facing copy, in API surfaces, in log messages, or in commit subjects. Internal Rust type names that pre-date this convention are grandfathered until they migrate.

BannedWhyUse instead
reactive (as in "reactive system" / "reactive view")Carries a negative connotation — sounds like firefighting rather than dataflow.live (for materialised / standing things), event-driven (for fire-on-event), trigger (for the binding itself), event bus (for pub/sub)
legacy in user-facing copyImplies neglect or shame.prior, pre-vN.M.K, or just drop the qualifier
broker (when talking about ArcFlow's bus)Implies a separate process you'd run alongside. ArcFlow's bus is in-process.event bus, in-process bus, bus
"reactive systems" in the taglineSame as reactive.live systems or event-driven systems or agent systems

What to call each new feature#

When shipping a new feature, name it from its layer:

  • A new manifest format, codec, or storage primitive → "World Store feature"
  • A new observation-time discipline rule or sensor-ingest path → "Perception Lake feature"
  • A new index, type, or constraint on typed entities → "World Graph feature"
  • A new planner rewrite, cost model, or executor capability → "Query Engine feature"
  • A new aggregator that runs in standing queries → "Live Surface feature"
  • A new pub/sub primitive (e.g. wildcard subscriptions) → "Event Bus feature"
  • A new behavior-tree node → "Behavior Engine primitive"
  • A new CALL procedure that does an algorithm → "Algorithm Library entry"

In commit subjects this looks like:

feat(core/worldstore): partition-pruning index in the manifest reader
feat(core/worldgraph): composite index for (label, position_prop)
feat(core/pubsub): NATS-compatible topic wildcard subscriptions
feat(core/algo): weighted_centroid fusion CALL procedure
feat(core/live): EMA aggregator over the Live Surface

Avoid genericised commits like feat(core): add stuff — the layer prefix tells the reader where in the architecture the change lives.

Cross-references#

  • Event Bus capability — the canonical surface for the Event Bus layer.
  • Live queries — the canonical surface for the Live Surface.
  • Algorithms — the canonical surface for the Algorithm Library.
  • Behavior graphs — the canonical surface for the Behavior Engine.
  • Architecture — operational detail for the whole binary.
← PreviousGlossaryNext →Data Types