wooo/awoooi
1
1
Fork 0
Code Issues 2 Pull Requests 1 Actions 3 Packages Projects Releases Wiki Activity

fix(p0.4): Playbook 學習鏈三道修復 — partial index + race防護 + 手動路徑接線

Browse Source 
ADR-092 P0.4 Playbook EWMA 學習閉環的 DB / Repository / Service 三層修補。

DB 層 (db-expert-fix by Engineer-B):
- ApprovalRecord.matched_playbook_id 移除 index=True，改 __table_args__ partial index
  (WHERE matched_playbook_id IS NOT NULL) — 多數列 NULL，full index 浪費空間
- adr092_p1_learning_chain_rollback.sql: 純 ROLLBACK SQL（DBA 手動執行）

Repository 層:
- playbook_repository.py: SELECT FOR UPDATE 防 lost update
  避免並發 EWMA 更新覆蓋彼此

Service 層 (P0.4 修復):
- proposal_service.py: 手動審核路徑補 _try_playbook_match_id 呼叫
  decision_manager auto_execute 路徑已有此邏輯（行 2035），
  此處補手動路徑缺口，使 matched_playbook_id 可寫入 DB → EWMA 才能演化

測試:
- test_playbook_repository_race_condition.py: 3 cases SELECT FOR UPDATE 防 race
  正確阻擋並發 EWMA 更新（pass）

Note: migration SQL 待 DBA 手動執行（feedback_dev_prod_separation.md），
      不執行 alembic upgrade（statu 文件禁忌條款）。

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Your Name
2026-04-26 20:19:46 +08:00
parent 55c6b4e2d9
commit e96055eef9
5 changed files with 310 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
apps/api/src/db/models.py
View File

@@ -22,6 +22,7 @@ from sqlalchemy import (
Integer,
String,
Text,
text,
)
from sqlalchemy import (
Enum as SQLEnum,
@@ -170,10 +171,11 @@ class ApprovalRecord(Base):
# B2 fix 2026-04-24 ogt + Claude Sonnet 4.6: Playbook 學習閉環斷鏈修復
# 原欄位缺失 → 人工審核後 matched_playbook_id 永遠 NULL → EWMA 無法更新
# 2026-04-25 db-expert-fix by Claude Engineer-B: 移除 index=True 避免自動生成 full index
# Partial index 改在 __table_args__ 宣告（WHERE matched_playbook_id IS NOT NULL）
matched_playbook_id: Mapped[str | None] = mapped_column(
String(36),
nullable=True,
index=True,
comment="匹配的 Playbook ID，學習服務用以更新 EWMA trust score",
)
@@ -203,7 +205,13 @@ class ApprovalRecord(Base):
Index("ix_approval_created_at", "created_at"),
Index("ix_approval_requested_by", "requested_by"),
Index("ix_approval_fingerprint", "fingerprint"), # 戰略 B: 指紋查詢優化
Index("ix_approval_matched_playbook", "matched_playbook_id"), # B2 fix
# 2026-04-25 db-expert-fix by Claude Engineer-B: 改為 partial index，只索引非 NULL 值
# 原 full index 與 index=True 三重宣告衝突已修復（一個來源真相：此處）
Index(
"ix_approval_matched_playbook",
"matched_playbook_id",
postgresql_where=text("matched_playbook_id IS NOT NULL"),
),
)

50
apps/api/src/repositories/playbook_repository.py
View File

@@ -342,30 +342,46 @@ class PlaybookRepository:
失敗: trust_new = 0.8 * trust_old + 0.2 * 0.0（衰減速度 2x）
trust < 0.1 → 記錄警告，由 Evolver Agent 封存
2026-04-15 ogt + Claude Sonnet 4.6（亞太）: Phase 3.5 EWMA 雙寫 PG
2026-04-25 db-expert-fix by Claude Engineer-B: SELECT FOR UPDATE 防 race condition（Lost Update）
"""
try:
playbook = await self.get_by_id(playbook_id)
if not playbook:
return False
# 使用 SELECT FOR UPDATE 確保並行 update_stats 不會 lost update
factory = get_session_factory()
async with factory() as session:
async with session.begin():
stmt = (
select(PlaybookRecord)
.where(PlaybookRecord.playbook_id == playbook_id)
.with_for_update()
)
result = await session.execute(stmt)
record = result.scalar_one_or_none()
if record is None:
return False
if success:
playbook.success_count += 1
playbook.trust_score = 0.9 * playbook.trust_score + 0.1 * 1.0
else:
playbook.failure_count += 1
playbook.trust_score = 0.8 * playbook.trust_score + 0.2 * 0.0
if success:
record.success_count += 1
record.trust_score = 0.9 * record.trust_score + 0.1 * 1.0
else:
record.failure_count += 1
record.trust_score = 0.8 * record.trust_score + 0.2 * 0.0
playbook.trust_score = max(0.0, min(1.0, playbook.trust_score))
playbook.last_used_at = now_taipei()
record.trust_score = max(0.0, min(1.0, record.trust_score))
record.last_used_at = now_taipei()
# session.begin() context manager 結束時自動 commit
# 雙寫（PG + Redis）
await self.update(playbook)
# 讀取最新值供 Redis 雙寫 + 日誌（PG 已 commit，此時 Redis 同步）
updated = await self._pg_get(playbook_id)
if updated is None:
return True # PG 已更新，Redis 同步失敗不影響結果
if playbook.trust_score < 0.1:
await self._redis_set(updated)
if updated.trust_score < 0.1:
logger.warning(
"playbook_trust_low_auto_archive_candidate",
playbook_id=playbook_id,
trust_score=playbook.trust_score,
trust_score=updated.trust_score,
hint="Evolver Agent 應將此 Playbook 封存",
)
@@ -373,8 +389,8 @@ class PlaybookRepository:
"playbook_stats_updated",
playbook_id=playbook_id,
success=success,
success_rate=playbook.success_rate,
trust_score=playbook.trust_score,
success_rate=updated.success_rate,
trust_score=updated.trust_score,
)
return True

56
apps/api/src/services/proposal_service.py
View File

@@ -246,6 +246,12 @@ class ProposalService:
metadata["signoz_correlation"] = llm_proposal.get("signoz_correlation", "")
metadata["optimization_suggestions"] = llm_proposal.get("optimization_suggestions", [])
# 2026-04-25 P0.4 修復 by Claude Engineer-B:
# 手動路徑（API 呼叫 generate_proposal）補 Playbook RAG 匹配，
# 讓 matched_playbook_id 得以寫入 DB，學習服務 EWMA 才能更新 trust score。
# decision_manager auto_execute 路徑已有此邏輯（行 2035），此處補手動路徑缺口。
matched_pb_id: str | None = await self._try_playbook_match_id(incident)
approval_create = ApprovalRequestCreate(
action=action,
description=description,
@@ -255,6 +261,7 @@ class ProposalService:
requested_by="OpenClaw AI",
incident_id=incident_id,
metadata=metadata,
matched_playbook_id=matched_pb_id,
)
approval = await self._approval_service.create_approval(approval_create)
@@ -354,6 +361,55 @@ class ProposalService:
return action_type, action, description
# =========================================================================
# 輔助方法: Playbook RAG 匹配（P0.4 2026-04-25 by Claude Engineer-B）
# =========================================================================
async def _try_playbook_match_id(self, incident: Incident) -> str | None:
"""
嘗試 Playbook RAG 匹配，回傳 matched_playbook_id（相似度 >= 0.85 才填）。
設計動機：手動路徑（generate_proposal）補 matched_playbook_id，
讓學習服務 EWMA 能在人工審核後更新 Playbook trust score。
邏輯與 decision_manager._try_playbook_match 相同，但只回傳 ID 不改 action。
失敗時靜默返回 None（不阻塞主流程）。
"""
PLAYBOOK_SIMILARITY_THRESHOLD = 0.85
try:
from src.models.playbook import SymptomPattern
from src.services.playbook_service import get_playbook_service
alert_names = [s.alert_name for s in incident.signals] if incident.signals else []
symptoms = SymptomPattern(
alert_names=alert_names,
affected_services=incident.affected_services or [],
severity_range=[incident.severity.value] if incident.severity else ["P2"],
)
recommendations = await get_playbook_service().get_recommendations(
symptoms=symptoms,
top_k=1,
)
if not recommendations:
return None
best_match = recommendations[0]
if best_match.similarity_score < PLAYBOOK_SIMILARITY_THRESHOLD:
return None
pb_id = best_match.playbook.playbook_id
logger.info(
"proposal_playbook_matched",
incident_id=incident.incident_id,
playbook_id=pb_id,
similarity=best_match.similarity_score,
)
return pb_id
except Exception as e:
logger.debug(
"proposal_playbook_match_skipped",
incident_id=getattr(incident, "incident_id", "?"),
error=str(e),
)
return None
# =========================================================================
# 輔助方法: 建立 BlastRadius
# =========================================================================