Metrics (metrics:) and business terms (@terms)

"What is revenue?" and "does the order count include cancelled orders?" are core domain facts. Defined separately in code, SQL and each BI tool, they give different numbers. Write them once in the DSL; the Apache Ossie export carries them to BI and AI tools as Ossie metrics. No application code or query API is generated from them.

model/metrics.yml (the pet store)

metrics:
  revenue:         sum(PurchaseOrder.total) where status != CANCELLED @terms(sales, 매출액) -- 매출 (취소 제외, VAT 포함·할인 후 금액)
  orderCount:      count(PurchaseOrder) where status != CANCELLED @terms(orders, 주문 건수) -- 주문 수 (취소 제외)
  averageOrder:    revenue / orderCount @terms(AOV, 객단가) -- 주문 1건 평균 금액
  shippedOrders:   count(PurchaseOrder) where status = SHIPPED -- 배송된 주문 수
  cancelledOrders: count(PurchaseOrder) where status = CANCELLED -- 취소된 주문 수
  customerCount:   count(Customer) where role = USER @terms(members, 회원 수) -- 고객 수 (관리자 제외)

(The descriptions after -- are the sample's own, in Korean: revenue without cancelled orders, VAT included and the discount taken off; orders without cancelled ones; the average order; shipped; cancelled; customers without administrators.)

metrics: is a top-level section in any model file of the domain. One line per measure: name: expression [@terms(...)] [-- description].

Syntax

Write	Means	Exported SQL (OSSIE_SQL_2026)
sum(Entity.field) / avg / min / max	an aggregate of one entity's field	SUM(PurchaseOrder.total)
count(Entity)	rows	COUNT(PurchaseOrder.order_id) - the key column, so a BI tool knows which dataset is counted
count(Entity.field)	rows with a value	COUNT(PurchaseOrder.total)
... where field != value and field >= value	a condition: =, !=, <, <=, >, >=, and	SUM(CASE WHEN ... THEN ... END)
values	numbers, 'text' (a ' inside written ''), true/false, enum constant names	the stored value: an @db code ('CAN'), the name with @jpa:Enumerated(STRING), otherwise the ordinal
metricA / metricB (+ - * /)	measures declared above it	(...) / NULLIF((...), 0) - a division never divides by zero; a count over a count is a decimal
@terms(sales, 매출액)	business terms and synonyms	ai_context.synonyms
-- revenue	the description	description

Data types: count is Integer; sum, min, max the field's type; avg and divisions Decimal (Float for a float field); + - * Integer when both sides are, otherwise Decimal. sum and avg take numeric fields only.

What the export wrote for revenue:

blueprint/semantic_model.yml (excerpt)

metrics:
  - name: revenue
    expression:
      dialects:
        - dialect: OSSIE_SQL_2026
          expression: "SUM(CASE WHEN PurchaseOrder.status <> 'CAN' THEN PurchaseOrder.total END)"
    description: "매출 (취소 제외, VAT 포함·할인 후 금액)"
    datatype: Decimal
    ai_context:
      synonyms: ["sales", "매출액"]

Limits

• One entity at a time. A measure that needs a condition on another entity ("items sold, counting only orders not cancelled" - an order line and its order's status) cannot be written yet. Do not force it: it would count the cancelled ones too.
• No time axis in a measure. Monthly revenue is the BI tool's: it slices by a date field (orderDate, Ossie's time dimension).
• A line ElastiCORE cannot read - an unknown entity or field, an enum constant the enum does not have, a measure declared below, an unquoted value for a field that is not an enum - is left out with a warning [ossie] [domain] metric name: ... ; it is left out. Nothing is guessed.
• A line with # or : in it must be quoted as a whole, or YAML cuts it or cannot read the file: ratio: "shippedOrders / orderCount -- shipped ratio: shipped / all".

@terms - the words people use

@terms on an entity's meta: line, a field or a measure lists what the business calls it. It does not change generated code; it goes to Ossie's ai_context.synonyms, so a question about "last month's purchase amount" leads an AI to PurchaseOrder and revenue.

PurchaseOrder:
  meta: entity @root @expose(20) @terms(order, purchase, 구매) -- 주문
Item:
  meta: entity @root @editable @terms(pet, 반려동물) -- 판매 품목
  fields:
    unitCost: money(10.2)! @terms(price, 가격)

An API that answers a measure

The metrics generate no endpoint. An administrator's summary is a method of your own in the owned service, behind ACCESS.require(currentUser, "report") - see the shop recipe's report in Your logic.

Checked: the pet store's MetricsTest places, ships and cancels orders, runs each exported metric's SQL on the database, and compares it with the same numbers computed from the entities.