Canonical CMMS schema

Siphon's healthcare-CMMS connectors all map different source shapes (OData, an Azure-APIM POST-query gateway, ServiceNow) onto one normalized model. That's the point: you integrate once, to the standard — @siphon/canonical (CANONICAL_VERSION = "0.1").

The model is derived from real connectors, not invented — every field is populated by at least one shipped connector (AIMS3, FSI, Nuvolo).

Entities

Two entities today: Asset (medical-equipment / clinical device) and WorkOrder. Each has a set of SCREAMING_SNAKE columns with a type (string | number | boolean | timestamp), a required flag, and aliases — the real per-connector variant names that satisfy the field.

import { CANONICAL_SCHEMA, conformance } from "@siphon/canonical";

// e.g. ASSET_ID with aliases ["ASSET_NUMBER", "NAME", "ASSET_TAG"]
CANONICAL_SCHEMA.Asset; // CanonicalField[]

Reconciliation is the value: ASSET_ID, ASSET_NUMBER, and NAME are the same concept across three sources; the canonical field carries the aliases so every one maps home.

Conformance

conformance(entity, producedFieldNames) scores a connector's output columns against the schema — matching by canonical name or a known alias:

conformance("Asset", ["ASSET_NUMBER", "DESCRIPTION", "SERIAL_NUMBER", "STATUS", "SEGMENT"]);
// → { coveragePct, requiredCoveragePct, covered, missingRequired, missingOptional, extra, version }

• covered — canonical fields the connector populates.
• missingRequired / missingOptional — canonical fields it doesn't.
• extra — source-specific columns with no canonical home (kept in the warehouse data JSONB, just not part of the standard).

The report is honest about gaps by design — e.g. FSI exposes no source-updated timestamp, so its Asset conformance shows a required miss on SOURCE_UPDATED_AT_UTC. Connectors target the canonical fields (see the add-a-connector guide), and a high conformance score is a sales artifact: this source is fully normalized.