OntologyL_Cypher 1.0: Chainable DSL for Graph Ontology Reasoning

OntologyL_Cypher 1.0 is the dedicated graph object chain language for the OntoL-Manage platform. It does not replace Cypher nor dictate reasoning order; instead it extracts graph objects as JSON views and uses chain semantics to perform filtering, computation, validation, and writes, which the inference engine then executes along ontology paths.

Core Design Principles

Ontology-native : directly operates on nodes, relationships, properties, and paths.

Chain-first : common graph operations are written on a single chain.

Explicit paths : path() supports shortest path, relationship priority, and strong binding.

Decidable conditions : condition fields must end with a comparison or boolean method.

Restricted writes : set, remove, detach, empty only allowed in execution fields.

Fail-closed : null, empty collection, exception, or unknown language all evaluate to false with a Chinese reason.

Unified receipt : every execution returns a program value, a Chinese semantic message, and the current object.

Language Capabilities

The language provides entry points n() for nodes and r() for relationships. Reading and filtering use get(), filter(), path(); aggregation uses sum(), avg(), count(), group_by(); writing uses set(), remove(), detach(), empty(). Parameters are written bare; only the JSON property dictionary for node creation retains double quotes.

Examples

Node/relationship entry: n(Person), r(R_CONTAINS) Current object direct read: get(qty) > 0, has(email) JSONPath subpath: get(.payload.items[0].qty) Path query: path(楼层->房间->测点, R_CONTAINS) Collection aggregation: sum, avg, count, group_by Safe write: set, remove, detach, empty (execution fields only)

From a Sentence to a Verifiable Reasoning Chain

Example Scenario: Pressure Vessel Material Compliance

Business rule : If the medium contains chlorine and the operating temperature is not below 60°C, the cylinder material must not use ordinary austenitic stainless steel.

Execution receipt: read current object → condition judgment → locate rule node → output conclusion and impact. Empty conditions do not mean pass; empty results, exceptions, or unknown languages all fail closed.

Designed for AI, Built for Audit

The goal is to let models generate stable expressions rather than free-form queries.

Metrics

29 registered chain methods

13 standard relationship types between entities

10 runtime prefix language entries

3 default LLM generation language categories

Generation Boundaries

LLMs default to generating only OntologyL_Cypher, ontoL_allen:, ontoL_geo:.

The remaining 7 prefixes are reserved for import compilation, existing configuration maintenance, or debugging.

Expressions enter a unified executor; they are not directly concatenated into SQL.

Write operations are only allowed in execution fields.

Audit Evidence

Program value: xxx_val Chinese message: xxx_msg Current object: current Reasoning card: edge_card Impact attribution: ontoL_impact_oneself / ontoL_impacted_by Same data, same expression → same explainable result.

High Performance and High Extensibility

Performance Design

Cypher pushdown : selection, filtering, sorting, aggregation, and traversal are pushed to the graph database.

Single-chain execution : one expression completes fetch, judgment, and aggregation, reducing intermediate results and round trips.

On-demand fetch : business data is read only after conditions pass, avoiding invalid queries.

Replica isolation : reasoning runs in a replica space; original data remains read-only.

Extensibility Dimensions

Method extension : new methods enter a unified registry and dispatch table; call style stays consistent.

Language extension : 10 prefix entries carry semantics, rules, interfaces, MCP, and spatiotemporal capabilities.

Scenario extension : the same expressions are reused across different entities, fields, and relationships.

Upgrade and Iteration

Language kernel upgrade : add methods, selectors, or return specs without changing business call habits.

Execution capability upgrade : integrate new data sources, MCP tools, or spatiotemporal capabilities.

Business model upgrade : when ontology and rules change, update expressions to reuse existing reasoning chains.

Performance, extensibility, and upgrades serve the same goal: keep business rules continuously executable.

Putting Ontology Capabilities into Real Business

Scenarios

Manufacturing quality : process parameters, material rules, quality gates; write "can/cannot use" as executable conditions.

Energy and equipment : floor → room → measurement point path aggregation; compute energy consumption and anomalies by area, equipment, time window.

Supply chain : supplier access, fulfillment progress, inventory and order constraints; conditions and actions unified in one chain.

Asset O&M : status, time-series, spatial relationships linkage; locate impact scope and output disposal suggestions.

Applicable Signals

Rules scattered in Excel, code, and oral experience.

Need AI to generate verifiable business judgments.

Graph data must be directly understood by business people.

Same problem needs stable reuse across projects.

Existing Cypher queries struggle to carry business semantics.

Getting Started with OntologyL_Cypher

Formal name : OntologyL_Cypher<br/> Product positioning : OntoL-Manage dedicated ontology reasoning language<br/> Capability scope : nodes, relationships, paths, aggregation, condition validation, execution write-back<br/> Release version : V1.0 / 2026.11

Three Steps to Adoption

Pick a rule : start with a high-frequency, decidable business rule.

Connect a chain : clarify entities, fields, relationships, and execution results.

Replicate to scenario : template the rule and reuse by entity and scenario.

OntologyL_Cypher lets ontology go from "viewable" to "executable", making every AI judgment have language, data, and a receipt.

OntoL-Manage · OntologyL_Cypher 1.0 · 2026.11

This manual describes the current release capabilities of OntoL-Manage; specific functions are subject to the actual deployed version.