labor-arbitration-system/backend/app/services/similar_cases_local.py

import re

from sqlalchemy.orm import Session

from app.models import CaseRecord


def _tokens(text: str) -> set[str]:
    if not text:
        return set()
    chunk = text[:8000]
    return set(re.findall(r"[\w\u4e00-\u9fff]+", chunk))


def _jaccard(a: set[str], b: set[str]) -> float:
    if not a or not b:
        return 0.0
    inter = len(a & b)
    union = len(a | b)
    return inter / union if union else 0.0


def find_similar(db: Session, exclude_case_id: int, query_text: str, limit: int = 5) -> list[dict]:
    """
    不依赖向量库：用词集合 Jaccard 与库里历史案件粗排相似度（适合毕业设计演示）。
    """
    q = _tokens(query_text)
    rows = db.query(CaseRecord).filter(CaseRecord.id != exclude_case_id).all()
    scored: list[tuple[float, CaseRecord]] = []
    for row in rows:
        s = _jaccard(q, _tokens(row.raw_text or ""))
        scored.append((s, row))
    scored.sort(key=lambda x: x[0], reverse=True)
    return [
        {"score": float(s), "case_id": row.id, "case_name": row.case_name}
        for s, row in scored[:limit]
        if s > 0
    ]