feat(scripts): Phase 26 補寫腳本 — 從 approval_records 反向建立 KM

225 筆歷史告警處理記錄全部補寫到 knowledge_entries (INCIDENT_CASE)

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

scripts/backfill_km_from_approvals.py

@@ -0,0 +1,107 @@
+#!/usr/bin/env python3
+"""
+Phase 26: 從 approval_records 歷史資料補寫 KM
+2026-04-06 ogt
+
+問題: 258 筆 Approval 記錄，只有 9 筆 KM entry，
+      大量告警處理歷史沒有沉澱到 Knowledge Base
+
+做法: 遍歷所有 APPROVED / EXECUTION_SUCCESS / EXECUTION_FAILED 的記錄，
+      建立 INCIDENT_CASE 類型的 KM entry (不重複)
+"""
+
+import asyncio
+import os
+import sys
+
+import asyncpg
+
+
+async def backfill():
+    db_url = os.environ.get("DATABASE_URL", "").replace("postgresql+asyncpg://", "postgresql://")
+    if not db_url:
+        print("ERROR: DATABASE_URL not set")
+        sys.exit(1)
+
+    conn = await asyncpg.connect(db_url)
+
+    # 取得所有有意義的 approval records (已批准/執行)
+    approvals = await conn.fetch("""
+        SELECT id, action, description, status, risk_level, created_at, incident_id
+        FROM approval_records
+        WHERE status IN ('APPROVED', 'EXECUTION_SUCCESS', 'EXECUTION_FAILED')
+        ORDER BY created_at ASC
+    """)
+
+    print(f"找到 {len(approvals)} 筆 Approval 記錄")
+
+    # 取得已存在的 KM entries（避免重複）
+    existing = await conn.fetch("""
+        SELECT related_approval_id FROM knowledge_entries
+        WHERE related_approval_id IS NOT NULL
+    """)
+    existing_ids = {r["related_approval_id"] for r in existing}
+    print(f"已有 {len(existing_ids)} 筆 KM entry 有 related_approval_id")
+
+    inserted = 0
+    skipped = 0
+
+    for ap in approvals:
+        ap_id = str(ap["id"])
+
+        if ap_id in existing_ids:
+            skipped += 1
+            continue
+
+        status = ap["status"]
+        status_icon = "✅" if status == "EXECUTION_SUCCESS" else ("⏳" if status == "APPROVED" else "❌")
+        status_text = {"EXECUTION_SUCCESS": "成功", "APPROVED": "已批准（未執行）", "EXECUTION_FAILED": "失敗"}[status]
+
+        title = f"[告警處理記錄] {status_icon} {(ap['action'] or '')[:60]}"
+        content = (
+            f"# {status_icon} 告警處理記錄\n\n"
+            f"**操作**: {ap['action'] or '無'}\n"
+            f"**說明**: {ap['description'] or '無'}\n"
+            f"**執行結果**: {status_text}\n"
+            f"**風險等級**: {ap['risk_level'] or '未知'}\n"
+            f"**Approval ID**: {ap_id}\n"
+            f"**Incident ID**: {ap['incident_id'] or '未關聯'}\n"
+            f"**建立時間**: {ap['created_at']}\n"
+        )
+
+        import json, uuid
+        tags_json = json.dumps(["backfill", "approval", status.lower()])
+        entry_id = str(uuid.uuid4())
+        await conn.execute("""
+            INSERT INTO knowledge_entries
+                (id, title, content, entry_type, category, tags, source, status,
+                 view_count, related_incident_id, related_approval_id,
+                 created_by, created_at, updated_at)
+            VALUES
+                ($1, $2, $3, 'INCIDENT_CASE', 'alert_handling',
+                 $4::json, 'AI_EXTRACTED', 'APPROVED',
+                 0, $5, $6, 'backfill_script', NOW(), NOW())
+        """,
+            entry_id,
+            title,
+            content,
+            tags_json,
+            ap["incident_id"],
+            ap_id,
+        )
+        inserted += 1
+
+    print(f"補寫完成: 新增 {inserted} 筆，略過 {skipped} 筆")
+
+    # 確認最終狀態
+    total = await conn.fetchrow("SELECT count(*) as cnt FROM knowledge_entries")
+    by_type = await conn.fetch("SELECT entry_type, count(*) FROM knowledge_entries GROUP BY entry_type ORDER BY count DESC")
+    print(f"\nKM 最終狀態 (共 {total['cnt']} 筆):")
+    for r in by_type:
+        print(f"  {r['entry_type']}: {r['count']}")
+
+    await conn.close()
+
+
+if __name__ == "__main__":
+    asyncio.run(backfill())