Data endpoints

Data endpoints are the named read and write points workflows use. A source node reads from one; a destination node writes to one; the ontology store is one. Integrations own the connection — endpoints name the usable doors. Connection side in Data and integrations.

Identity fields

Field	Meaning	Rules
Name	Human reference workflows point at	Stable — renaming breaks every node naming it
Namespace	Grouping, usually the system or domain	Keeps `crm.orders` distinct from `warehouse.orders`
Description	What the door is for, who uses it	Write for the next operator

{
  "name": "crm.orders",
  "namespace": "crm",
  "description": "Order stream for renewal analytics. Owner: revenue ops."
}

Image: endpoint list showing name, namespace, role, and health.

Role and contract

Field	Values	Meaning
Role	source, destination, ontology store, event source/sink, API, file store, search index, graph store, other	Which node kinds may use it
Contract kind	stream read, query read, plain read, write, API call, event subscribe/publish, ontology fact store, transactional table	The access pattern — match need to kind
Target	At least one selector field	The concrete table, path, stream, or object
Accepts / emits	Data formats in and out	Emits default analytics columnar; change only when consumers demand

{
  "name": "warehouse.forecast_output",
  "namespace": "warehouse",
  "role": "destination",
  "contractKind": "write",
  "target": { "schema": "analytics", "table": "forecast_scores" },
  "emits": "parquet"
}

Schema hints

Field	Meaning	Example
Columns	Expected fields — documentation for mappers, not enforcement	`["order_id", "total", "placed_at"]`
Primary key	Row identifiers feeding downstream identity	`["order_id"]`, or `["account_id", "snapshot_date"]`

Trust and ownership

Field	Values	Meaning
Trust rank	Low / medium / high	How much downstream work should lean on it; review on schedule
Owner team	Team name	Who fixes the door when it breaks
Freshness	Expected currency	The promise staleness is judged against, e.g. "hourly"

Forecasts built on low-trust doors deserve extra skepticism at dry-run.

Write governance

Field	Meaning
Write policy	Rules on mutations, identity, duplicates, conflicts, retention

Full catalog in Write policies. A destination without one is a door without rules — do not commit it.

Choosing well

One endpoint per real door — share endpoints, version workflows.
Namespaces mirror systems, not teams: systems outlive reorgs.
Trust ranks get reviewed on a schedule, or every door quietly becomes medium and the rank means nothing.