refactor(api): Phase 22 技術債修復 - 業務邏輯分層

P2.3: LearningService.get_learning_summary() 業務邏輯移至 Service 層
- Repository 只提供原始統計數據
- Service 計算 best_action 和 learning_status

P2.6: Playbook similarity 計算邏輯抽取
- 新增 src/utils/similarity.py
- Repository 從 utils 導入，不再定義演算法

2026-03-31 Claude Code (首席架構師技術債修復)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

apps/api/src/repositories/playbook_repository.py

@@ -11,6 +11,9 @@ Phase 7.2: Repository 實作
 - 實作 IPlaybookRepository Protocol
 - Redis 為 Working Memory (7天 TTL)
 - PostgreSQL 為 Episodic Memory
+
+Phase 22 P2: 相似度計算邏輯移至 utils
+2026-03-31 Claude Code (首席架構師技術債修復)
 """
 
 import json
@@ -24,6 +27,7 @@ from src.models.playbook import (
     SymptomPattern,
 )
 from src.repositories.interfaces import IPlaybookRepository
+from src.utils.similarity import calculate_symptom_similarity
 from src.utils.timezone import now_taipei
 
 logger = structlog.get_logger(__name__)
@@ -37,63 +41,6 @@ PLAYBOOK_INDEX_ALERT_PREFIX = "playbook:index:alert:"
 PLAYBOOK_INDEX_SERVICE_PREFIX = "playbook:index:service:"
 
 
-def _calculate_jaccard_similarity(set_a: set, set_b: set) -> float:
-    """計算 Jaccard 相似度"""
-    if not set_a and not set_b:
-        return 1.0
-    intersection = len(set_a & set_b)
-    union = len(set_a | set_b)
-    return intersection / union if union > 0 else 0.0
-
-
-def calculate_symptom_similarity(
-    pattern_a: SymptomPattern,
-    pattern_b: SymptomPattern,
-) -> float:
-    """
-    計算症狀相似度
-
-    算法: 加權 Jaccard 相似度
-
-    維度權重:
-    - alert_names: 0.35 (最重要)
-    - affected_services: 0.30
-    - severity: 0.15
-    - keywords: 0.20
-
-    Returns:
-        float: 0.0 ~ 1.0 相似度分數
-    """
-    weights = {
-        "alert_names": 0.35,
-        "affected_services": 0.30,
-        "severity": 0.15,
-        "keywords": 0.20,
-    }
-
-    scores = {
-        "alert_names": _calculate_jaccard_similarity(
-            set(pattern_a.alert_names),
-            set(pattern_b.alert_names),
-        ),
-        "affected_services": _calculate_jaccard_similarity(
-            set(pattern_a.affected_services),
-            set(pattern_b.affected_services),
-        ),
-        "severity": (
-            1.0
-            if set(pattern_a.severity_range) & set(pattern_b.severity_range)
-            else 0.0
-        ),
-        "keywords": _calculate_jaccard_similarity(
-            set(pattern_a.keywords),
-            set(pattern_b.keywords),
-        ),
-    }
-
-    return sum(weights[k] * scores[k] for k in weights)
-
-
 class PlaybookRepository:
     """
     Playbook Repository 實作

apps/api/src/services/learning_service.py

@@ -655,7 +655,12 @@ class LearningService:
         """
         取得學習摘要
 
-        2026-03-29 P0 修正: 委託 Repository 實作
+        Phase 22 P2: 業務邏輯移至 Service 層
+        2026-03-31 Claude Code (首席架構師技術債修復)
+
+        邏輯:
+        - 從 Repository 取得原始統計數據
+        - 在 Service 層計算 best_action 和 learning_status
 
         Returns:
             {
@@ -667,7 +672,51 @@ class LearningService:
                 'learning_status': 'sufficient',
             }
         """
-        return await self._repository.get_learning_summary(anomaly_key)
+        # 從 Repository 取得原始統計
+        all_stats = await self._repository.get_all_repair_stats(anomaly_key)
+
+        if not all_stats:
+            return {
+                "anomaly_key": anomaly_key,
+                "total_repair_attempts": 0,
+                "overall_success_rate": 0.0,
+                "actions_tried": [],
+                "best_action": None,
+                "learning_status": "insufficient",
+            }
+
+        # === 以下為業務邏輯，應在 Service 層 ===
+
+        total_attempts = sum(s["total"] for s in all_stats.values())
+        total_success = sum(s["success"] for s in all_stats.values())
+        overall_rate = total_success / total_attempts if total_attempts > 0 else 0.0
+
+        # 找出最佳動作 (需要至少 3 次數據)
+        best_action = None
+        best_rate = 0.0
+        for action, stats in all_stats.items():
+            if stats["total"] >= 3 and stats["success_rate"] > best_rate:
+                best_rate = stats["success_rate"]
+                best_action = {"action": action, "success_rate": best_rate}
+
+        # 判斷學習狀態
+        if total_attempts < 3:
+            learning_status = "insufficient"
+        elif total_attempts < 10:
+            learning_status = "learning"
+        elif overall_rate >= 0.8:
+            learning_status = "excellent"
+        else:
+            learning_status = "sufficient"
+
+        return {
+            "anomaly_key": anomaly_key,
+            "total_repair_attempts": total_attempts,
+            "overall_success_rate": overall_rate,
+            "actions_tried": list(all_stats.keys()),
+            "best_action": best_action,
+            "learning_status": learning_status,
+        }
 
     def _get_action_tier(self, action: str) -> int:
         """取得動作的 Tier"""

apps/api/src/utils/similarity.py

@@ -0,0 +1,87 @@
+"""
+Similarity Calculation Utils
+=============================
+Phase 22 P2: 將相似度計算邏輯從 Repository 移出
+
+設計原則:
+- 演算法邏輯應獨立於資料存取層
+- Repository 只負責 CRUD，不負責演算法
+- Service 層可以使用這些工具函數
+
+版本: v1.0
+建立: 2026-03-31 (台北時區)
+建立者: Claude Code (首席架構師技術債修復)
+"""
+
+from src.models.playbook import SymptomPattern
+
+
+def calculate_jaccard_similarity(set_a: set, set_b: set) -> float:
+    """
+    計算 Jaccard 相似度
+
+    Jaccard = |A ∩ B| / |A ∪ B|
+
+    Args:
+        set_a: 集合 A
+        set_b: 集合 B
+
+    Returns:
+        float: 0.0 ~ 1.0
+    """
+    if not set_a and not set_b:
+        return 1.0  # 兩個空集合視為完全相同
+    if not set_a or not set_b:
+        return 0.0
+
+    intersection = len(set_a & set_b)
+    union = len(set_a | set_b)
+    return intersection / union if union > 0 else 0.0
+
+
+def calculate_symptom_similarity(
+    pattern_a: SymptomPattern,
+    pattern_b: SymptomPattern,
+) -> float:
+    """
+    計算症狀相似度
+
+    算法: 加權 Jaccard 相似度
+
+    維度權重:
+    - alert_names: 0.35 (最重要)
+    - affected_services: 0.30
+    - severity: 0.15
+    - keywords: 0.20
+
+    Returns:
+        float: 0.0 ~ 1.0 相似度分數
+    """
+    weights = {
+        "alert_names": 0.35,
+        "affected_services": 0.30,
+        "severity": 0.15,
+        "keywords": 0.20,
+    }
+
+    scores = {
+        "alert_names": calculate_jaccard_similarity(
+            set(pattern_a.alert_names),
+            set(pattern_b.alert_names),
+        ),
+        "affected_services": calculate_jaccard_similarity(
+            set(pattern_a.affected_services),
+            set(pattern_b.affected_services),
+        ),
+        "severity": (
+            1.0
+            if set(pattern_a.severity_range) & set(pattern_b.severity_range)
+            else 0.0
+        ),
+        "keywords": calculate_jaccard_similarity(
+            set(pattern_a.keywords),
+            set(pattern_b.keywords),
+        ),
+    }
+
+    return sum(weights[k] * scores[k] for k in weights)