feat(dotnet): add normalization contract harness
CI Workflow Lint / validate-ci-workflow-lint (push) Failing after 14s
Validators (Pushes and Pull Requests) / validate-ui-and-storage (push) Successful in 20s
Validators (Pushes and Pull Requests) / validate-core (push) Has been cancelled

This commit is contained in:
2026-07-13 00:31:13 +09:00
parent b694a101d1
commit ed137c2574
7 changed files with 223 additions and 0 deletions
+1
View File
@@ -1471,6 +1471,7 @@ WBS-8.8 (KIS 리팩터) — 독립적 (원격 병행)
> parity 계약: [WBS_10_DOTNET_PARITY_CONTRACT.yaml](./WBS_10_DOTNET_PARITY_CONTRACT.yaml)
> provenance 계약: [WBS_10_DOTNET_PROVENANCE_CONTRACT.yaml](./WBS_10_DOTNET_PROVENANCE_CONTRACT.yaml)
> scheduler contract: [WBS_10_DOTNET_SCHEDULER_CONTRACT.yaml](./WBS_10_DOTNET_SCHEDULER_CONTRACT.yaml)
> normalization contract: [WBS_10_DOTNET_NORMALIZATION_CONTRACT.yaml](./WBS_10_DOTNET_NORMALIZATION_CONTRACT.yaml)
> 현황 진단(2026-06-26): .NET 프로젝트는 Python 엔진(41 모듈, 14,500 LOC) 대비 5~10%(~1,400 LOC) 수준.
> Domain 계산기 6개·데이터 모델 8개·KIS/Naver/Yahoo 클라이언트·PostgreSQL 마이그레이션·Razor Pages 어드민 대시보드 기본 구현 완료.
@@ -0,0 +1,85 @@
formula_id: WBS_10_DOTNET_NORMALIZATION_CONTRACT_V1
owner: QuantEngine
status: draft
goal: "쓰기 경로 정규화와 읽기 경로 역정규화 경계를 고정한다."
canonical_write_path:
schema: engine_history
tables:
- source_observation
- factor_definition
- factor_observation
- decision_event
- decision_factor_evidence
- outcome_evaluation
invariant:
- "source_observation은 원천 관측 1건당 1행"
- "factor_definition은 (factor_id, factor_version) 단일 원장"
- "factor_observation은 observation_id를 반드시 참조"
- "decision_event는 decision_key로 단일 식별"
- "outcome_evaluation은 decision_id + horizon_days 조합으로 단일 식별"
canonical_read_path:
view: engine_history.training_example_v1
purpose: "모델 학습/캘리브레이션용 역정규화 projection"
consumers:
- model_training
- calibration_jobs
- diagnostics
forbidden_patterns:
- "읽기 모델을 쓰기 원장으로 사용"
- "원천 payload를 읽기 projection에 중복 저장"
- "직렬 UI 조회를 위해 원장 테이블을 직접 조인해 장기 유지"
expected_fields:
normalized_tables:
source_observation:
- observation_id
- observed_at
- instrument_id
- source_name
- source_version
- payload
- provenance
factor_observation:
- factor_observation_id
- observation_id
- factor_id
- factor_version
- observed_at
- numeric_value
- text_value
- gate
- provenance
decision_event:
- decision_id
- decision_key
- decided_at
- instrument_id
- action
- gate
- score
- source_version
- trace
- provenance
denormalized_view:
- decision_id
- decision_key
- decided_at
- instrument_id
- action
- decision_gate
- score
- source_version
- horizon_days
- realized_return
- benchmark_return
- excess_return
- outcome_class
- evaluation_gate
- factor_features
notes:
- "정규화는 쓰기 중복 제거와 provenance 보존이 목적이다."
- "역정규화는 학습/진단 편의용 projection으로만 허용한다."