feat: QYLAW 法律法规知识库

- 语义检索(FAISS + embedding)+ 精确查找(法规名+条号) - RAG 问答(SSE 流式,支持 thinking 折叠显示) - 法规浏览(原文阅读) - 历史记录(检索+对话持久化到 SQLite) - 设置页(系统提示词/模板/LLM 参数可配置) - 检索质量评估脚本 Generated with [Devin](https://devin.ai) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-08-07 14:55:25 +08:00
commit 641e33b834
39 changed files with 5254 additions and 0 deletions
@@ -0,0 +1,89 @@
+"""检索 API — /api/search"""
+from fastapi import APIRouter, Request, HTTPException, Query
+from typing import Optional
+
+from app.services.retriever import Retriever
+from app.services.embedding import embed_text
+from app.services.history import add_search_history
+from app.config import MAX_QUERY_LENGTH
+
+router = APIRouter()
+
+
+@router.get("/search")
+async def search(
+    request: Request,
+    query: str = Query(..., min_length=1, max_length=MAX_QUERY_LENGTH, description="查询文本"),
+    mode: str = Query("semantic", description="检索模式:semantic(语义) / keyword(关键词精确)"),
+    category: Optional[str] = Query(None, description="法规类别过滤"),
+    province: Optional[str] = Query(None, description="省份过滤"),
+    city: Optional[str] = Query(None, description="市级过滤"),
+    top_k: int = Query(20, ge=1, le=100, description="返回数量"),
+    page: int = Query(1, ge=1, description="页码"),
+    page_size: int = Query(20, ge=1, le=100, description="每页数量"),
+):
+    """检索法规条文(语义检索或关键词精确搜索)
+
+    - mode=semantic: 向量化查询 → FAISS 检索 → metadata 过滤(默认)
+    - mode=keyword: 按法规名+条号精确匹配,支持"郑州市劳动用工条例 第三十二条"
+    """
+    trace_id = getattr(request.state, "trace_id", "")
+
+    retriever = Retriever.get_instance()
+    if not retriever.is_ready():
+        raise HTTPException(status_code=503, detail="索引未加载,请等待或检查索引文件")
+
+    if mode == "keyword":
+        # 关键词精确搜索
+        results = retriever.keyword_search(
+            query=query,
+            top_k=top_k * 3 if (category or province or city) else top_k,
+            category=category,
+            province=province,
+            city=city,
+        )
+    else:
+        # 语义检索
+        try:
+            query_vec = await embed_text(query)
+        except Exception as e:
+            raise HTTPException(
+                status_code=503,
+                detail=f"embedding 服务不可用: {str(e)}",
+            )
+        fetch_k = top_k * 3 if (category or province or city) else top_k
+        results = retriever.search(
+            query_vec=query_vec,
+            top_k=fetch_k,
+            category=category,
+            province=province,
+            city=city,
+        )
+
+    # 分页
+    total = len(results)
+    start = (page - 1) * page_size
+    end = start + page_size
+    page_results = results[start:end]
+
+    # 记录检索历史
+    add_search_history(
+        query=query,
+        category=category,
+        province=province,
+        city=city,
+        result_count=total,
+        top_k=top_k,
+    )
+
+    return {
+        "code": 0,
+        "message": "ok",
+        "data": {
+            "results": page_results,
+            "total": total,
+            "page": page,
+            "page_size": page_size,
+        },
+        "trace_id": trace_id,
+    }