TECHNICAL ARCHITECTURE

The deterministic layer around probabilistic AI.

Axiomyx keeps identity, structure, relationships and provenance outside the model so evidence can persist across agents, providers and actions.

EVIDENCE PIPELINE

Canonicalise once. Retrieve through a stable contract.

01

Canonical object contract

Selected source content and metadata enter one deterministic representation with source identity and time retained.

02

ZCM geometry

Reproducible multi-metric structure creates a stable signature and geometric position outside the model.

03

PostgreSQL storage

Customer-controlled PostgreSQL stores canonical objects, indexes, provenance and federated relationships.

04

Federation

Supported connectors keep source systems separate while comparable canonical records share one indexed relationship layer.

05

Graph provenance

Explicit source, transformation, version and relationship edges reconstruct the evidence path.

06

Deterministic replay

The canonical object, structural signature and provenance contract can be inspected and replayed independently of model wording.

07

Entitlement architecture

Signed local entitlement governs edition and deployment scope without embedding a signing private key in customer packages.

08

Activation boundary

Activation exchanges bounded licence and installation metadata, not customer operational data.

MODEL AND CONNECTOR BOUNDARIES

Integration and evidence are separate concerns.

A parser or connector handles source-specific extraction and mapping. Axiomyx handles deterministic canonical identity, ZCM geometry, indexing, federation and provenance after the selected information reaches the canonical contract.

The customer is not limited to an Axiomyx-only connector catalogue. Customer-built adapters and existing extraction pipelines can use the same versioned canonical input contract.

Source systems remain authoritative and only selected data enters the evidence layer. Connector availability remains release-specific. External model selection remains customer-controlled.

Source systems remain authoritative → selected records become canonical objects → ZCM supplies structural identity → PostgreSQL preserves the evidence layer → authorised AI retrieves compact evidence.