fix: 修复35个安全与功能缺陷，补全知识进化/数字孪生/行为采集模块
## 安全修复 (12项) - Webhook接口添加全局Token认证，过滤敏感请求头 - 修复JWT Base64 padding公式，防止签名验证绕过 - 数据库密码/飞书Token从源码移除，改为环境变量 - 工作流引擎添加路径遍历防护 (_resolve_safe_path) - eval()添加模板长度上限检查 - 审批API添加认证依赖 - 前端v-html增强XSS转义，console.log仅开发模式输出 - 500错误不再暴露内部异常详情 ## Agent运行时修复 (7项) - 删除_inject_knowledge_context中未定义db变量的finally块 - 工具执行添加try/except保护，异常不崩溃Agent - LLM重试计入budget计数器 - self_review异常时passed=False - max_iterations截断标记success=False - 工具参数JSON解析失败时记录警告日志 - run()开始时重置_llm_invocations计数器 ## 配置与基础设施 - DEBUG默认False，SQL_ECHO独立配置项 - init_db()补全13个缺失模型导入 - 新增WEBHOOK_AUTH_TOKEN/SQL_ECHO配置项 - 新增.env.example模板文件 ## 前端修复 (12项) - 登录改用URLSearchParams替代FormData - 401拦截器通过Pinia store统一清理状态 - SSE流超时从60s延长至300s - final/error事件时清除streamTimeout - localStorage聊天记录添加24h TTL - safeParseArgCount替代模板中裸JSON.parse - fetchUser 401时同时清除user对象 ## 新增模块 - 知识进化: knowledge_extractor/retriever/tasks - 数字孪生: shadow_executor/comparison模型 - 行为采集: behavior_middleware/collector/fingerprint_engine - 代码审查: code_review_agent/document_review_agent - 反馈学习: feedback_learner - 瓶颈检测/优化引擎/成本估算/需求估算 - 速率限制器 (rate_limiter) - Alembic迁移 015-020 ## 文档 - 商业化落地计划 - 8篇docs文档 (架构/API/部署/开发/贡献等) - Docker Compose生产配置 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
renjianbo
168
backend/app/services/optimization_engine.py
Normal file
168
backend/app/services/optimization_engine.py
Normal file
@@ -0,0 +1,168 @@
|
||||
"""
|
||||
工作流自动优化引擎 — 根据瓶颈检测结果自动生成优化方案
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
import logging
|
||||
from typing import Any, Dict, List, Optional
|
||||
|
||||
from app.services.bottleneck_detector import bottleneck_detector
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class OptimizationEngine:
|
||||
"""工作流自动优化引擎 — 生成 DAG 优化方案"""
|
||||
|
||||
def analyze_and_optimize(self, hours: int = 24) -> Dict[str, Any]:
|
||||
"""运行完整分析并生成优化方案。"""
|
||||
analysis = bottleneck_detector.run_full_analysis(hours=hours)
|
||||
bottlenecks = analysis.get("bottlenecks", [])
|
||||
optimizations = self.generate_optimizations(bottlenecks)
|
||||
dag_changes = self.generate_dag_changes(optimizations)
|
||||
|
||||
return {
|
||||
"period_hours": hours,
|
||||
"summary": analysis,
|
||||
"optimizations": optimizations,
|
||||
"dag_changes": dag_changes,
|
||||
"requires_approval": len(optimizations) > 0,
|
||||
}
|
||||
|
||||
def generate_optimizations(self, bottlenecks: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
|
||||
"""根据瓶颈生成具体优化方案。"""
|
||||
optimizations = []
|
||||
|
||||
for b in bottlenecks:
|
||||
severity = b.get("severity", "low")
|
||||
if severity == "low":
|
||||
continue
|
||||
|
||||
opt = {
|
||||
"node_type": b["node_type"],
|
||||
"severity": severity,
|
||||
"current_metrics": {
|
||||
"p95_ms": b.get("p95_ms"),
|
||||
"error_rate": b.get("error_rate"),
|
||||
"p50_ms": b.get("p50_ms"),
|
||||
},
|
||||
"changes": [],
|
||||
}
|
||||
|
||||
if b.get("is_bottleneck"):
|
||||
opt["changes"].append({
|
||||
"type": "parallelize",
|
||||
"description": f"将 {b['node_type']} 拆分为并行子节点",
|
||||
"estimated_improvement": "延迟降低 50-70%",
|
||||
"implementation": {
|
||||
"action": "replace_node",
|
||||
"new_structure": "parallel_gateway",
|
||||
"sub_nodes": self._suggest_split(b["node_type"]),
|
||||
},
|
||||
})
|
||||
|
||||
if b.get("is_problematic"):
|
||||
opt["changes"].append({
|
||||
"type": "add_validation",
|
||||
"description": f"添加前置校验节点，失败率 {b['error_rate']:.1%}",
|
||||
"estimated_improvement": "失败率降至 5% 以下",
|
||||
"implementation": {
|
||||
"action": "insert_before",
|
||||
"new_node": {
|
||||
"type": "validate",
|
||||
"name": f"PreValidate_{b['node_type']}",
|
||||
"config": {"required_fields": [], "timeout_ms": 5000},
|
||||
},
|
||||
},
|
||||
})
|
||||
opt["changes"].append({
|
||||
"type": "add_retry",
|
||||
"description": "添加重试逻辑：最多 3 次，指数退避",
|
||||
"estimated_improvement": "间歇性失败自动恢复",
|
||||
"implementation": {
|
||||
"action": "update_config",
|
||||
"config": {"retry_count": 3, "retry_delay_ms": 1000, "backoff": "exponential"},
|
||||
},
|
||||
})
|
||||
|
||||
if b.get("is_inefficient"):
|
||||
opt["changes"].append({
|
||||
"type": "add_cache",
|
||||
"description": f"添加结果缓存，TTL: 300s",
|
||||
"estimated_improvement": "重复调用减少 30-60%",
|
||||
"implementation": {
|
||||
"action": "insert_before",
|
||||
"new_node": {
|
||||
"type": "cache_check",
|
||||
"name": f"Cache_{b['node_type']}",
|
||||
"config": {"ttl_seconds": 300, "key_fields": []},
|
||||
},
|
||||
},
|
||||
})
|
||||
|
||||
if opt["changes"]:
|
||||
optimizations.append(opt)
|
||||
|
||||
return optimizations
|
||||
|
||||
def _suggest_split(self, node_type: str) -> List[Dict[str, Any]]:
|
||||
"""根据节点类型建议拆分方案。"""
|
||||
suggestions = {
|
||||
"llm_call": [
|
||||
{"name": "PromptPreprocess", "type": "transform"},
|
||||
{"name": "LLMCall_Fast", "type": "llm_call", "config": {"model": "fast"}},
|
||||
{"name": "LLMCall_Accurate", "type": "llm_call", "config": {"model": "accurate"}},
|
||||
{"name": "ResultMerge", "type": "merge"},
|
||||
],
|
||||
"api_request": [
|
||||
{"name": "RequestPrepare", "type": "transform"},
|
||||
{"name": "APICall_Core", "type": "api_request"},
|
||||
{"name": "ResponseParse", "type": "transform"},
|
||||
],
|
||||
"data_query": [
|
||||
{"name": "QueryBuilder", "type": "transform"},
|
||||
{"name": "DBQuery", "type": "data_query"},
|
||||
{"name": "ResultFormat", "type": "transform"},
|
||||
],
|
||||
"code_execution": [
|
||||
{"name": "SandboxSetup", "type": "setup"},
|
||||
{"name": "CodeRun", "type": "code_execution"},
|
||||
{"name": "OutputParse", "type": "transform"},
|
||||
],
|
||||
}
|
||||
return suggestions.get(node_type, [
|
||||
{"name": "PreProcess", "type": "transform"},
|
||||
{"name": "CoreExecute", "type": "execute"},
|
||||
{"name": "PostProcess", "type": "transform"},
|
||||
])
|
||||
|
||||
def generate_dag_changes(self, optimizations: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
|
||||
"""将优化方案转换为 DAG 变更列表（可用于 diff 展示）。"""
|
||||
changes = []
|
||||
for opt in optimizations:
|
||||
for change in opt.get("changes", []):
|
||||
impl = change.get("implementation", {})
|
||||
changes.append({
|
||||
"node_type": opt["node_type"],
|
||||
"severity": opt["severity"],
|
||||
"action": impl.get("action"),
|
||||
"description": change["description"],
|
||||
"estimated_improvement": change.get("estimated_improvement"),
|
||||
"detail": impl,
|
||||
})
|
||||
return changes
|
||||
|
||||
def apply_optimization(self, workflow_id: str, optimization_id: str,
|
||||
approved_changes: List[str]) -> Dict[str, Any]:
|
||||
"""应用用户确认的优化变更（创建新版本）。"""
|
||||
# 这里与 workflow_version 系统集成，创建优化后的新版本
|
||||
# 当前返回框架结果，实际集成在后续完成
|
||||
return {
|
||||
"workflow_id": workflow_id,
|
||||
"status": "pending",
|
||||
"message": f"优化方案已记录，{len(approved_changes)} 项变更待应用",
|
||||
"applied_changes": approved_changes,
|
||||
}
|
||||
|
||||
|
||||
optimization_engine = OptimizationEngine()
|
||||
Reference in New Issue
Block a user