Included vs Custom

What's Included vs. What Requires Customization

Included, working as installed

• 182 data models with 4,560 fields, deployed and queryable
• 465 forward relations and 283 reverse collections, enforced by the platform
• Provenance columns on every model
• Source-key deduplication
• Identity derivation for 21 keyless lookup tables
• A computed load order (7 layers) published per endpoint
• An endpoint switchboard listing all 182, shipped disabled

Included, but needs a value only you can supply

Item	What is needed	Consequence if skipped
ExternalSystem record	A code identifying this Plex instance	Every insert refused
Site identity	PCN and source-system id	Extraction flows refuse to run
Enabled endpoints	Which objects this site reads	Nothing loads
Extraction flows	A separate package	Schema stays empty
Plex API connection	Credentials for this PCN	No reads

Requires custom development

Item	Why
40 multi-level sub-resources	Their URL carries more than one path parameter and no single parent supplies them all. These flows refuse to run rather than requesting a URL with a brace in it.
Cross-PCN or cross-ERP harmonisation	This is a spoke. Canonical entities across sites belong to the enterprise hub.
Writing to Plex	Read mirror by design.
Screens, dashboards, reports	No UI is included.
Business logic on landed data	Scheduling, MRP, quality rules — the schema is a foundation, not an application.
Endpoints Plex adds after 2026-08-05	Regenerate from a newer dictionary; the generator makes this cheap, but it is a build step.

Two Honest Gaps

Six models have no reliable key. plexNote is free text that genuinely repeats; five others are settings blocks carrying only numbers and JSON, whose real identity is a parent record they do not yet reference. Their externalId falls back to row position, which survives a re-read of an unchanged list and nothing more. They are shipped disabled and flagged rather than given a fabricated key.

Plex documents no pagination. Across 283 read contracts there is no limit/offset, no cursor and no continuation token — _limit appears on two endpoints and $top on one, and both truncate rather than page. 21 endpoints accept a created/modified date window, which is the only real way to bound a large extract and the basis for incremental sync. The remaining 158 return what they return. This is a property of the vendor's API, not of the accelerator, and no amount of packaging changes it.