mini-openclaw-v2/backend/data_layer/history_store.py

"""历史评审案例向量库

功能：
1. 存储历史评审案例（项目特征+评审结果+评审员行为）
2. 相似案例检索（基于向量相似度）
3. 评审员行为学习（团队集体经验，非个人预测）
"""

import json
import os
from typing import List, Dict, Optional, Tuple
from datetime import datetime
from pathlib import Path
import logging

logger = logging.getLogger(__name__)


class HistoryStore:
    """历史评审案例存储（简化版，生产环境建议使用 Chroma/Qdrant）"""

    def __init__(self, data_dir: str = "./xingbang_data/history"):
        self.data_dir = Path(data_dir)
        self.data_dir.mkdir(parents=True, exist_ok=True)
        self.cases_file = self.data_dir / "review_cases.jsonl"

        # 确保文件存在
        if not self.cases_file.exists():
            self.cases_file.touch()

    def add_case(self, case_data: Dict) -> str:
        """添加一条历史评审案例"""
        case_id = f"case_{datetime.now().strftime('%Y%m%d%H%M%S')}"

        case_record = {
            "case_id": case_id,
            "timestamp": datetime.now().isoformat(),
            **case_data
        }

        # 追加写入 JSONL
        with open(self.cases_file, "a", encoding="utf-8") as f:
            f.write(json.dumps(case_record, ensure_ascii=False) + "\n")

        logger.info(f"案例已保存: {case_id}")
        return case_id

    def search_similar_cases(
        self,
        industry: str,
        credit_amount: float,
        risk_level: str = None,
        limit: int = 5
    ) -> List[Dict]:
        """检索相似案例（简化版：基于行业+金额范围）"""

        if not self.cases_file.exists():
            return []

        similar_cases = []

        with open(self.cases_file, "r", encoding="utf-8") as f:
            for line in f:
                if not line.strip():
                    continue

                try:
                    case = json.loads(line)

                    # 行业匹配
                    industry_match = industry in case.get("industry", "") or case.get("industry", "") in industry

                    # 金额范围匹配（±30%）
                    case_amount = case.get("credit_amount", 0)
                    amount_match = 0.7 * credit_amount <= case_amount <= 1.3 * credit_amount

                    # 风险等级匹配（可选）
                    risk_match = risk_level is None or case.get("risk_level") == risk_level

                    if industry_match and amount_match and risk_match:
                        similar_cases.append(case)

                except json.JSONDecodeError:
                    continue

        # 按时间倒序返回最近的案例
        similar_cases.sort(key=lambda x: x.get("timestamp", ""), reverse=True)
        return similar_cases[:limit]

    def get_case_by_id(self, case_id: str) -> Optional[Dict]:
        """根据 ID 获取案例详情"""

        if not self.cases_file.exists():
            return None

        with open(self.cases_file, "r", encoding="utf-8") as f:
            for line in f:
                if not line.strip():
                    continue

                try:
                    case = json.loads(line)
                    if case.get("case_id") == case_id:
                        return case
                except json.JSONDecodeError:
                    continue

        return None

    def get_team_patterns(self, industry: str = None) -> Dict:
        """提取团队集体经验模式（高频风险点、常见处理方式）"""

        if not self.cases_file.exists():
            return {"high_frequency_risks": [], "common_solutions": []}

        risk_counter = {}
        solution_counter = {}

        with open(self.cases_file, "r", encoding="utf-8") as f:
            for line in f:
                if not line.strip():
                    continue

                try:
                    case = json.loads(line)

                    # 行业过滤
                    if industry and industry not in case.get("industry", ""):
                        continue

                    # 统计高频风险点
                    for risk in case.get("identified_risks", []):
                        risk_key = risk.get("type", "unknown")
                        risk_counter[risk_key] = risk_counter.get(risk_key, 0) + 1

                    # 统计常见解决方案
                    solution = case.get("solution", "")
                    if solution:
                        solution_counter[solution] = solution_counter.get(solution, 0) + 1

                except json.JSONDecodeError:
                    continue

        # 排序
        high_frequency_risks = sorted(
            [{"risk": k, "count": v} for k, v in risk_counter.items()],
            key=lambda x: x["count"],
            reverse=True
        )[:10]

        common_solutions = sorted(
            [{"solution": k, "count": v} for k, v in solution_counter.items()],
            key=lambda x: x["count"],
            reverse=True
        )[:10]

        return {
            "high_frequency_risks": high_frequency_risks,
            "common_solutions": common_solutions,
        }

    def get_statistics(self) -> Dict:
        """获取统计数据"""

        if not self.cases_file.exists():
            return {"total_cases": 0}

        total_cases = 0
        industry_dist = {}
        approval_rate = {"approved": 0, "rejected": 0, "pending": 0}

        with open(self.cases_file, "r", encoding="utf-8") as f:
            for line in f:
                if not line.strip():
                    continue

                try:
                    case = json.loads(line)
                    total_cases += 1

                    # 行业分布
                    industry = case.get("industry", "未知")
                    industry_dist[industry] = industry_dist.get(industry, 0) + 1

                    # 通过率
                    result = case.get("review_result", "pending")
                    if result in approval_rate:
                        approval_rate[result] += 1

                except json.JSONDecodeError:
                    continue

        return {
            "total_cases": total_cases,
            "industry_distribution": industry_dist,
            "approval_rate": approval_rate,
        }


# 全局单例
_history_store: Optional[HistoryStore] = None


def get_history_store() -> HistoryStore:
    """获取历史案例存储单例"""
    global _history_store
    if _history_store is None:
        _history_store = HistoryStore()
    return _history_store