Control Plane

Builds a semantic understanding of the enterprise by combining an ontology layer, a knowledge graph, and a vector embedding index — the brain that gives AI the context to reason about your business.

For platform-overview framing see Platform Architecture > Control Plane. For retrieval-engineering principles see AI Operating System > Semantic RAG.

Capabilities

• Ontology Schemas — YAML-defined business entity definitions (customers, orders, assets, contracts).
• Policies — business rules, guardrails, and compliance constraints.
• Workflows — multi-step business process definitions.
• Integrations — system-to-system sync mappings.
• Knowledge Graph — Apache AGE on PostgreSQL for entity relationships and graph traversal.
• Semantic Index — pgvector on PostgreSQL for hybrid (vector + graph) search.
• Hybrid retrieval — combine semantic similarity and graph traversal in a single query.

Architecture

graph TB
    subgraph Ontology["Ontology Bundle"]
        Schemas["Schemas<br/>Entity definitions"]
        Policies["Policies<br/>Business rules"]
        Workflows["Workflows<br/>Process definitions"]
        Integrations["Integrations<br/>System mappings"]
    end

    subgraph Storage["Storage Layer"]
        AGE["Apache AGE<br/>Knowledge Graph"]
        PGV["pgvector<br/>Semantic Embeddings"]
    end

    subgraph Query["Query Layer"]
        Hybrid["Hybrid Search<br/>Vector + Graph"]
        Explorer["Entity Explorer"]
        Editor["Knowledge Editor"]
    end

    Ontology --> AGE
    Ontology --> PGV
    AGE --> Hybrid
    PGV --> Hybrid
    Hybrid --> Explorer
    Hybrid --> Editor

Console Walkthrough

Six tabs cover ingestion, retrieval, governance, and orchestration of the semantic layer.

Knowledge Pipeline

• Real-time graph and embedding counters.
• Six-stage pipeline: Extract Unified Schema → Resolve FDW Mapping → Extract Mapping Rules → Extract Policies → Extract Workflows → Extract Integrations.
• Trigger manually or schedule.

Knowledge Index

• Filter chips swap between Schemas, Policies, Workflows, Integrations.
• Document detail shows metadata plus the canonical YAML body that feeds embeddings.

Knowledge Graph

• Nodes colour-coded by source system.
• Typed edges (triggers, syncs_to, depends_on, constrained_by, owned_by).
• Full-text search and per-source filtering.

Policy Builder

• Per-policy header: name, ID, risk level, target systems.
• Source-of-truth field binding.
• Editable Constraints, Approval Matrix, and Restrictions blocks.

Workflows

• Per-workflow header, trigger definition, steps, and dependencies.
• Cross-system orchestration as first-class.

Integration Rules

• Per-rule header: source/target system, frequency, direction (one-way or bidirectional).
• Field mappings, failure handling, dependencies in collapsible sections.

Related References

• Semantic RAG — retrieval engineering principles (canonical knowledge model, hybrid retrieval, black-box validation).
• Semantic RAG Pipeline — the production pipeline implementation.
• Implementation Guide — production code mapping (schema_indexer.py, vector_store.py, embedding_service.py, rag_service.py).
• Data Plane — the data layer this plane consumes.
• Reasoning Plane — the layer that consumes Control Plane retrievals.