The Knowledge Plane: A unified view of business operations

A Knowledge Graph (KG) describes a concrete graph data model with business entities, relationships and properties. Unlike SQL, Knowledge Graphs model relationships as first-class citizens, which supports efficient multi-hop traversals (i.e., Friends-of-a-Friend), queries and analytics over graph data.

Typically, KGs organize information around a specific application, database, report, integration, or analytical use-case. The result is a siloed KG, which is developer friendly, but not necessarily re-usable or applicable enterprise-wide. In EnterpriseWeb, KG entities are mapped to higher-level concepts, types and policies in EnterpriseWeb’s connected ontology, ensuring semantic interoperability across the two areas. For developers it means they can rapidly, declaratively model use cases that all reference shared terms and are under common governance.

​In addition, EnterpriseWeb's KG supports the modeling of complex objects, not just entities. The typed, immutable objects (i.e., “things, not strings”) can capture properties, behaviors, constraints, dependencies, interfaces, and configurations.

EnterpriseWeb's KG includes unified graph memory of all objects in a formal system catalog and inventory. The catalog includes version control and change history, and the inventory provides lifecycle management and configuration history.

The append-only, log-style memory captures the ever-changing state and telemetry of every object not just to maintain dashboard views, but as machine-readable triggers for actions. It turns the KG into a reactive database that supports service lifecycle management and continuous assurance over streaming data.

EnterpriseWeb’s KG enables change data capture for audit, troubleshooting, root cause analysis, NetOps automation and Agentic AI. In addition, the log-style memory retains temporal history, which supports predictive analytics and learning for proactive maintenance and process optimization.

For more powerful inferences, customers need a formal conceptual model; an Ontology.

Ontologies describe classes, relationships and axioms (i.e., assertions and rules about a domain). They model high-level tribal knowledge about a domain de-coupled from any specific application. By abstracting historic silos Ontologies enable a "shared understanding" across teams, technologies and use cases enabling end-to-end reasoning. Related terms with disparate labels in the underlying data can map up to common concepts in the ontology to promote interoperability, while preserving nuanced distinctions in definitions and implementations.

The third leg of EnterpriseWeb’s domain modeling is the Topology. The Topology models business operations as a graph of nodes representing the interrelationships of physical and virtual entities (e.g., services, resources, and clusters).

The topology represents a digital twin of business operations. It is mapped to concepts in the ontology and rich objects in the KG’s inventory. The Topology supports flexible views that can be traversed, filtered and mapped all the way down to objects in KG inventory with state and telemetry. It allows Ops teams and agents to rapidly identify the source of an issue and its graph dependencies for troubleshooting, root-cause analysis and AIOps.

Each area of the Knowledge Plane is a logically partitioned dimension of a single graph, which supports efficient system-wide retrieval, queries, analytics, inferences and constraint satisfaction problems. Entities in the knowledge graph are mapped to the ontology, which extends reasoning all the way down to the graph data. Likewise, nodes in the topology reference the ontology for common business meaning and map to objects in the KG for shared state and telemetry.

The Semantic AI Harness enforces common governance over the ontology, topology and KG with native security/identity, reliable messaging, transaction management, and state management. It includes a platform abstraction (i.e., metamodel, type system and libraries), which provide primitives, building blocks and services that ease and speed the design, use and management of the Knowledge Plane.

EnterpriseWeb makes it easy to model complex, real-world domains. It allows customers to import files, models and documents to bootstrap development of the ontology, topology and KG. EnterpriseWeb also supports zero-copy integration. It can introspect existing data sources (i.e., SQL, GQL and SPARQL) and system endpoints (REST, CLI, etc.). EnterpriseWeb dynamically extracts entities, algorithmically maps and types them to generate domain models and objects.

EnterpriseWeb’s no-code platform provides a model-driven process for onboarding distributed solution elements, federated components, virtualized data sources and physical devices. Customers can import files, docs, models and reference public endpoints. EnterpriseWeb introspects the information, performs algorithmic mapping, entity extraction, dynamic typing and generates a typed, immutable object under management.

EnterpriseWeb performs exact match typing based on its platform-level metamodel and type system, as well as any industry or customer domain-specific types. If information supplied is unclear or unknown to EnterpriseWeb or the domain, it automatically performs adaptive type inference and returns a scored/ranked type. In either case, users are prompted to confirm the proposed type or manually re-map before proceeding (i.e., Humans-in-the-Loop).

Once verified, EnterpriseWeb generates a type-specific schema, auto-fills properties and wraps the object with universal resources indicator (URI) over a set of attached interfaces. This design-time step is a form of minimum viable normalization that supports global discovery, declarative composition and policy-controlled management. EnterpriseWeb supports policy-based validations and automates conformance checking. In this way, EnterpriseWeb ensures well-formed typed objects over diverse, multi-vendor solution elements, components, data sources and devices.

Note: EnterpriseWeb is not based on static, hierarchical class-based inheritance as is found in Object Oriented Programming (OOP), as well as Semantic Web and Graph Databases. Class-based inheritance forces square pegs into round holes and doesn’t support real-world variety found across objects of the same type. Instead, EnterpriseWeb uses Prototypal Inheritance, which provides flexibility when modeling heterogeneous objects with proprietary interfaces, properties and behaviors.

EnterpriseWeb’s common object model supports three tiers of information per type: 1) standard schema and values; 2) standard schemas and non-standard values (vendor/item specific); and 3) unstructured, schemaless information in its raw form.

The common object accommodates structured, semi-structured and unstructured information in a single typed object with a universal resource identifier (URI) so that searches, queries and inferences can rapidly and precisely identify candidate objects and drill down into the tiers of information as necessary for optimal response. EnterpriseWeb’s runtime realizes the complex objects based on interaction context.

EnterpriseWeb’s approach is ideal for working with language models and agents as it provides just enough structure for deterministic discovery, while supporting probabilistic inferences over fuzzy information.

EnterpriseWeb is a knowledge-driven workflow orchestrator that supports integration, virtualization and federation by design. It lets EnterpriseWeb to leverage external data sources and services as part of a declaratively modeled complex event.

• Data Service Integration: EnterpriseWeb can introspect and model any mainstream protocol/format (REST, gRPC, JSON, YAML, CSV, XML, etc.) to create an adaptor object mapped to the ontology.
• Zero-copy Data Source Virtualization: The platform abstraction layer encodes the syntax and semantics for SQL, GQL and SPARQL query languages. EnterpriseWeb virtualizes the access to specific Relational databases, Graph databases and triple stores as modeled objects. On connection, the platform introspects the data sources model/schema and generates a set of objects representing the underlying entities, relationships and data to enable federated queries.
• Federate SQL, GQL and SPARQL Queries: EnterpriseWeb’s platform abstraction (i.e., metamodel, type system and libraries), includes modeled adaptors for SQL, GQL and SPARQL. It enables EnterpriseWeb to dynamically generate well-formed queries in the native syntax of a virtualized data source or materialized view that can be executed at runtime as part of a broad evaluation of conditions.
• Data Migration: Extract and transform concepts, objects/entities, relationships and rules from Semantic Web (.rdf, .nt, .ttl), Labeled Property Graph (GQL) and Relational DB (SQL).
• Model Extraction: Extract concepts, objects/entities, relationships and rules from other semi-structured models in REST, JSON, YAML, CSV, XML, etc.

EnterpriseWeb is a no-code platform for declaratively modeling complex operational domains, workflows and services; not an IDE for custom development of algorithms. That being said, EnterpriseWeb is open for extension. It supports code objects, which are modeled in EnterpriseWeb and can be persisted locally (i.e., system library) or in a repo (e.g., GitHub) and are executed by a federated engine.

The typed Code Objects can be referenced in EnterpriseWeb process models and configured per use case. At runtime, EnterpriseWeb resolves references and related policies, dynamically generating a schema and data set based on interaction context, and invokes the federated engine to execute the algorithm. Results are then used as additional context in the processing of a system event, request or prompt to optimize decisions and actions.

• Graph Data Science Libraries (e.g., Neo4J, TigerGraph, etc.): Allows the incorporation of neighboring and clustering algorithms, which Labelled Property Graphs are well suited for.
• SQL Analytics Libraries (e.g., Postgres, Relational AI, Observe, etc.): - Enables EnterpriseWeb to leverage SQL-based predictive analytics (i.e., centrality, similarity, community detection algorithms).
• Semantic Views (i.e., Snowflake OSI, or similar): EnterpriseWeb offers pre-integration with Snowflake and can generate and maintain YAML files for OSI-based Semantic Views, so customers can leverage real-time, always up-to-date SQL-based Natural Language analysis of the Semantic AI Harness and its use of Agents and Language Models

EnterpriseWeb can also orchestrate federated rules engines and classifiers. In addition to modeling logic directly in EnterpriseWeb, developers can invoke external rules engines, which can be modeled as objects and composed in workflows. The outputs can be used as inputs in a more complex decision.

Likewise, EnterpriseWeb supports orchestration of “System One” classifiers for fast, local typed decisions, open-model scoring, and task-specific training (e.g., Type Safe AI Jev, AWS Strands Decider, Cloudflare Clef, custom ML/RL algorithms and AI runtimes) for targeted use cases as part of a broader deterministic evaluation of conditions with domain knowledge, typed objects, business rules, workflow logic and active state/telemetry.

By bridging knowledge, data and process, EnterpriseWeb optimizes existing graph use cases and extends them beyond retrieval, queries and inferences to ontology-driven constraint satisfaction, behavioral interfaces, workflow automation and intent-based orchestration using a single language.