fix: derive all aggregation timeframes from trade timestamp (#112)

run_all_aggregations() previously used datetime.now(UTC) for weekly
through annual layers while using the trade date only for daily,
causing data misalignment on backfill. Now all layers consistently
use the latest trade timestamp. Also adds "Z" suffix handling for
fromisoformat() compatibility and strengthens test assertions to
verify L4-L2 layer values end-to-end.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

docs/requirements-log.md

@@ -64,25 +64,3 @@
 **참고:** Realtime 모드 전용. Daily 모드는 배치 효율성을 위해 정적 watchlist 사용.
 
 **이슈/PR:** #76, #77
-
----
-
-## 2026-02-10
-
-### 코드 리뷰 시 플랜-구현 일치 검증 규칙
-
-**배경:**
-- 코드 리뷰 시 플랜(EnterPlanMode에서 승인된 계획)과 실제 구현이 일치하는지 확인하는 절차가 없었음
-- 플랜과 다른 구현이 리뷰 없이 통과될 위험
-
-**요구사항:**
-1. 모든 PR 리뷰에서 플랜-구현 일치 여부를 필수 체크
-2. 플랜에 없는 변경은 정당한 사유 필요
-3. 플랜 항목이 누락되면 PR 설명에 사유 기록
-4. 스코프가 플랜과 일치하는지 확인
-
-**구현 결과:**
-- `docs/workflow.md`에 Code Review Checklist 섹션 추가
-  - Plan Consistency (필수), Safety & Constraints, Quality, Workflow 4개 카테고리
-
-**이슈/PR:** #114

docs/workflow.md

@@ -74,37 +74,3 @@ task_tool(
 ```
 
 Use `run_in_background=True` for independent tasks that don't block subsequent work.
-
-## Code Review Checklist
-
-**CRITICAL: Every PR review MUST verify plan-implementation consistency.**
-
-Before approving any PR, the reviewer (human or agent) must check ALL of the following:
-
-### 1. Plan Consistency (MANDATORY)
-
-- [ ] **Implementation matches the approved plan** — Compare the actual code changes against the plan created during `EnterPlanMode`. Every item in the plan must be addressed.
-- [ ] **No unplanned changes** — If the implementation includes changes not in the plan, they must be explicitly justified.
-- [ ] **No plan items omitted** — If any planned item was skipped, the reason must be documented in the PR description.
-- [ ] **Scope matches** — The PR does not exceed or fall short of the planned scope.
-
-### 2. Safety & Constraints
-
-- [ ] `src/core/risk_manager.py` is unchanged (READ-ONLY)
-- [ ] Circuit breaker threshold not weakened (only stricter allowed)
-- [ ] Fat-finger protection (30% max order) still enforced
-- [ ] Confidence < 80 still forces HOLD
-- [ ] No hardcoded API keys or secrets
-
-### 3. Quality
-
-- [ ] All new/modified code has corresponding tests
-- [ ] Test coverage >= 80%
-- [ ] `ruff check src/ tests/` passes (no lint errors)
-- [ ] No `assert` statements removed from tests
-
-### 4. Workflow
-
-- [ ] PR references the Gitea issue number
-- [ ] Feature branch follows naming convention (`feature/issue-N-description`)
-- [ ] Commit messages are clear and descriptive

src/context/aggregator.py

@@ -230,21 +230,44 @@ class ContextAggregator:
             )
 
     def run_all_aggregations(self) -> None:
-        """Run all aggregations from L7 to L1 (bottom-up)."""
+        """Run all aggregations from L7 to L1 (bottom-up).
+
+        All timeframes are derived from the latest trade timestamp so that
+        past data re-aggregation produces consistent results across layers.
+        """
+        cursor = self.conn.execute("SELECT MAX(timestamp) FROM trades")
+        row = cursor.fetchone()
+        if not row or row[0] is None:
+            return
+
+        ts_raw = row[0]
+        if ts_raw.endswith("Z"):
+            ts_raw = ts_raw.replace("Z", "+00:00")
+        latest_ts = datetime.fromisoformat(ts_raw)
+        trade_date = latest_ts.date()
+        date_str = trade_date.isoformat()
+
+        iso_year, iso_week, _ = trade_date.isocalendar()
+        week_str = f"{iso_year}-W{iso_week:02d}"
+        month_str = f"{trade_date.year}-{trade_date.month:02d}"
+        quarter = (trade_date.month - 1) // 3 + 1
+        quarter_str = f"{trade_date.year}-Q{quarter}"
+        year_str = str(trade_date.year)
+
         # L7 (trades) → L6 (daily)
-        self.aggregate_daily_from_trades()
+        self.aggregate_daily_from_trades(date_str)
 
         # L6 (daily) → L5 (weekly)
-        self.aggregate_weekly_from_daily()
+        self.aggregate_weekly_from_daily(week_str)
 
         # L5 (weekly) → L4 (monthly)
-        self.aggregate_monthly_from_weekly()
+        self.aggregate_monthly_from_weekly(month_str)
 
         # L4 (monthly) → L3 (quarterly)
-        self.aggregate_quarterly_from_monthly()
+        self.aggregate_quarterly_from_monthly(quarter_str)
 
         # L3 (quarterly) → L2 (annual)
-        self.aggregate_annual_from_quarterly()
+        self.aggregate_annual_from_quarterly(year_str)
 
         # L2 (annual) → L1 (legacy)
         self.aggregate_legacy_from_annual()

tests/test_context.py

@@ -300,9 +300,17 @@ class TestContextAggregator:
         # Verify data exists in each layer
         store = aggregator.store
         assert store.get_context(ContextLayer.L6_DAILY, date, "total_pnl") == 1000.0
-        current_week = datetime.now(UTC).strftime("%Y-W%V")
-        assert store.get_context(ContextLayer.L5_WEEKLY, current_week, "weekly_pnl") is not None
-        # Further layers depend on time alignment, just verify no crashes
+        from datetime import date as date_cls
+        trade_date = date_cls.fromisoformat(date)
+        iso_year, iso_week, _ = trade_date.isocalendar()
+        trade_week = f"{iso_year}-W{iso_week:02d}"
+        assert store.get_context(ContextLayer.L5_WEEKLY, trade_week, "weekly_pnl") is not None
+        trade_month = f"{trade_date.year}-{trade_date.month:02d}"
+        trade_quarter = f"{trade_date.year}-Q{(trade_date.month - 1) // 3 + 1}"
+        trade_year = str(trade_date.year)
+        assert store.get_context(ContextLayer.L4_MONTHLY, trade_month, "monthly_pnl") == 1000.0
+        assert store.get_context(ContextLayer.L3_QUARTERLY, trade_quarter, "quarterly_pnl") == 1000.0
+        assert store.get_context(ContextLayer.L2_ANNUAL, trade_year, "annual_pnl") == 1000.0
 
 
 class TestLayerMetadata: