2026-04-21 11:02:16 +08:00
|
|
|
|
"""
|
2026-05-04 02:01:22 +08:00
|
|
|
|
Qdrant 混合检索器模块
|
2026-04-21 11:02:16 +08:00
|
|
|
|
|
2026-05-04 02:01:22 +08:00
|
|
|
|
提供基于 Qdrant 的混合检索(Dense + Sparse)功能,包括:
|
|
|
|
|
|
- 纯混合检索(无子父文档)
|
|
|
|
|
|
- 父子文档混合检索(先检索子文档,再返回父文档)
|
2026-04-21 11:02:16 +08:00
|
|
|
|
|
|
|
|
|
|
核心原理:
|
2026-05-04 02:01:22 +08:00
|
|
|
|
- 使用 Qdrant 原生 Fusion API (RRF) 做分数融合
|
|
|
|
|
|
- 同时使用稠密向量(语义)和稀疏向量(BM25 关键词)
|
2026-04-21 11:02:16 +08:00
|
|
|
|
"""
|
|
|
|
|
|
|
2026-05-04 02:01:22 +08:00
|
|
|
|
from typing import Dict, Any, Optional, List
|
2026-04-21 11:02:16 +08:00
|
|
|
|
from qdrant_client import QdrantClient
|
|
|
|
|
|
from qdrant_client.http.exceptions import UnexpectedResponse
|
2026-05-04 02:01:22 +08:00
|
|
|
|
from qdrant_client.http.models import (
|
|
|
|
|
|
SearchRequest, Fusion, FusionProtocol, NamedVector, NamedSparseVector
|
2026-05-03 17:58:21 +08:00
|
|
|
|
)
|
2026-05-04 02:01:22 +08:00
|
|
|
|
from langchain_core.documents import Document
|
2026-04-21 11:02:16 +08:00
|
|
|
|
from langchain_core.embeddings import Embeddings
|
2026-05-04 02:01:22 +08:00
|
|
|
|
from langchain_core.retrievers import BaseRetriever, RetrieverOutput
|
2026-04-21 11:02:16 +08:00
|
|
|
|
|
2026-05-04 02:01:22 +08:00
|
|
|
|
from rag_core import QdrantVectorStore, get_sparse_embedder, create_docstore
|
2026-04-21 19:06:34 +08:00
|
|
|
|
from rag_core.client import create_qdrant_client as create_core_qdrant_client
|
2026-04-29 10:52:01 +08:00
|
|
|
|
from app.model_services import get_embedding_service
|
2026-05-04 02:01:22 +08:00
|
|
|
|
from app.logger import info, warning, debug
|
|
|
|
|
|
|
2026-04-21 11:02:16 +08:00
|
|
|
|
|
|
|
|
|
|
# 模块级常量
|
|
|
|
|
|
DEFAULT_SEARCH_K = 20
|
2026-05-04 02:01:22 +08:00
|
|
|
|
DEFAULT_PARENT_SEARCH_K = 5
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class HybridRetriever(BaseRetriever):
|
|
|
|
|
|
"""
|
|
|
|
|
|
混合检索器:稠密向量 + BM25 稀疏向量 RRF 分数融合
|
|
|
|
|
|
|
|
|
|
|
|
直接使用 Qdrant 原生 Fusion API,性能最优。
|
|
|
|
|
|
"""
|
|
|
|
|
|
|
|
|
|
|
|
def __init__(
|
|
|
|
|
|
self,
|
|
|
|
|
|
collection_name: str,
|
|
|
|
|
|
vector_store: QdrantVectorStore,
|
|
|
|
|
|
search_k: int = DEFAULT_SEARCH_K,
|
|
|
|
|
|
):
|
|
|
|
|
|
"""
|
|
|
|
|
|
Args:
|
|
|
|
|
|
collection_name: Qdrant 集合名称
|
|
|
|
|
|
vector_store: QdrantVectorStore 实例
|
|
|
|
|
|
search_k: 检索返回结果数
|
|
|
|
|
|
"""
|
|
|
|
|
|
self.collection_name = collection_name
|
|
|
|
|
|
self.vector_store = vector_store
|
|
|
|
|
|
self.search_k = search_k
|
|
|
|
|
|
self.client = vector_store.get_qdrant_client()
|
|
|
|
|
|
self.sparse_embedder = get_sparse_embedder()
|
|
|
|
|
|
|
|
|
|
|
|
def _get_relevant_documents(
|
|
|
|
|
|
self, query: str, *, run_manager: Optional[Any] = None
|
|
|
|
|
|
) -> List[Document]:
|
|
|
|
|
|
"""
|
|
|
|
|
|
同步检索相关文档
|
|
|
|
|
|
|
|
|
|
|
|
Args:
|
|
|
|
|
|
query: 查询字符串
|
|
|
|
|
|
run_manager: LangChain 运行管理器(可选)
|
|
|
|
|
|
|
|
|
|
|
|
Returns:
|
|
|
|
|
|
相关文档列表
|
|
|
|
|
|
"""
|
|
|
|
|
|
# 生成双向量
|
|
|
|
|
|
dense_query = self.vector_store.embeddings.embed_query(query)
|
|
|
|
|
|
sparse_query = self.sparse_embedder.embed_query(query)
|
|
|
|
|
|
|
|
|
|
|
|
# 构建双检索请求
|
|
|
|
|
|
searches = [
|
|
|
|
|
|
# 稠密检索
|
|
|
|
|
|
SearchRequest(
|
|
|
|
|
|
vector=NamedVector(name="dense", vector=dense_query),
|
|
|
|
|
|
limit=self.search_k,
|
|
|
|
|
|
with_payload=True
|
|
|
|
|
|
),
|
|
|
|
|
|
# 稀疏检索
|
|
|
|
|
|
SearchRequest(
|
|
|
|
|
|
vector=NamedSparseVector(name="sparse", vector=sparse_query),
|
|
|
|
|
|
limit=self.search_k,
|
|
|
|
|
|
with_payload=True
|
|
|
|
|
|
)
|
|
|
|
|
|
]
|
|
|
|
|
|
|
|
|
|
|
|
# RRF 分数融合
|
|
|
|
|
|
fused_results = self.client.fusion(
|
|
|
|
|
|
collection_name=self.collection_name,
|
|
|
|
|
|
requests=searches,
|
|
|
|
|
|
fusion=Fusion(fusion=FusionProtocol.RRF)
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
# 转换为 Document 格式
|
|
|
|
|
|
results = []
|
|
|
|
|
|
for point in fused_results.points:
|
|
|
|
|
|
doc = Document(
|
|
|
|
|
|
page_content=point.payload.pop("text", ""),
|
|
|
|
|
|
metadata=point.payload
|
|
|
|
|
|
)
|
|
|
|
|
|
results.append(doc)
|
|
|
|
|
|
|
|
|
|
|
|
debug(f"混合检索返回 {len(results)} 个文档")
|
|
|
|
|
|
return results
|
|
|
|
|
|
|
|
|
|
|
|
async def _aget_relevant_documents(
|
|
|
|
|
|
self, query: str, *, run_manager: Optional[Any] = None
|
|
|
|
|
|
) -> List[Document]:
|
|
|
|
|
|
"""异步检索(当前调用同步版本)"""
|
|
|
|
|
|
# Qdrant 客户端没有原生 async,这里用同步版本
|
|
|
|
|
|
return self._get_relevant_documents(query, run_manager=run_manager)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class ParentHybridRetriever(BaseRetriever):
|
|
|
|
|
|
"""
|
|
|
|
|
|
父子文档混合检索器:
|
|
|
|
|
|
|
|
|
|
|
|
1. 先用混合检索找到相关子文档
|
|
|
|
|
|
2. 根据子文档的 parent_id 找到对应的父文档
|
|
|
|
|
|
3. 去重并返回父文档
|
|
|
|
|
|
"""
|
|
|
|
|
|
|
|
|
|
|
|
def __init__(
|
|
|
|
|
|
self,
|
|
|
|
|
|
collection_name: str,
|
|
|
|
|
|
vector_store: QdrantVectorStore,
|
|
|
|
|
|
search_k: int = DEFAULT_PARENT_SEARCH_K,
|
|
|
|
|
|
docstore: Optional[Any] = None,
|
|
|
|
|
|
):
|
|
|
|
|
|
"""
|
|
|
|
|
|
Args:
|
|
|
|
|
|
collection_name: Qdrant 集合名称
|
|
|
|
|
|
vector_store: QdrantVectorStore 实例
|
|
|
|
|
|
search_k: 最终返回的父文档数
|
|
|
|
|
|
docstore: 文档存储(如果父文档在 PostgreSQL),可选
|
|
|
|
|
|
"""
|
|
|
|
|
|
self.collection_name = collection_name
|
|
|
|
|
|
self.vector_store = vector_store
|
|
|
|
|
|
self.search_k = search_k
|
|
|
|
|
|
self.client = vector_store.get_qdrant_client()
|
|
|
|
|
|
self.sparse_embedder = get_sparse_embedder()
|
|
|
|
|
|
self.docstore = docstore
|
|
|
|
|
|
|
|
|
|
|
|
def _get_relevant_documents(
|
|
|
|
|
|
self, query: str, *, run_manager: Optional[Any] = None
|
|
|
|
|
|
) -> List[Document]:
|
|
|
|
|
|
"""
|
|
|
|
|
|
同步检索相关父文档
|
|
|
|
|
|
|
|
|
|
|
|
Args:
|
|
|
|
|
|
query: 查询字符串
|
|
|
|
|
|
run_manager: LangChain 运行管理器(可选)
|
|
|
|
|
|
|
|
|
|
|
|
Returns:
|
|
|
|
|
|
相关父文档列表
|
|
|
|
|
|
"""
|
|
|
|
|
|
# 1. 生成查询双向量
|
|
|
|
|
|
dense_query = self.vector_store.embeddings.embed_query(query)
|
|
|
|
|
|
sparse_query = self.sparse_embedder.embed_query(query)
|
|
|
|
|
|
|
|
|
|
|
|
# 2. 多取一些子文档,避免去重后数量不足
|
|
|
|
|
|
search_limit = self.search_k * 2
|
|
|
|
|
|
searches = [
|
|
|
|
|
|
# 稠密检索
|
|
|
|
|
|
SearchRequest(
|
|
|
|
|
|
vector=NamedVector(name="dense", vector=dense_query),
|
|
|
|
|
|
limit=search_limit,
|
|
|
|
|
|
with_payload=True
|
|
|
|
|
|
),
|
|
|
|
|
|
# 稀疏检索
|
|
|
|
|
|
SearchRequest(
|
|
|
|
|
|
vector=NamedSparseVector(name="sparse", vector=sparse_query),
|
|
|
|
|
|
limit=search_limit,
|
|
|
|
|
|
with_payload=True
|
|
|
|
|
|
)
|
|
|
|
|
|
]
|
|
|
|
|
|
|
|
|
|
|
|
# 3. RRF 分数融合,拿到子文档命中结果
|
|
|
|
|
|
fused_results = self.client.fusion(
|
|
|
|
|
|
collection_name=self.collection_name,
|
|
|
|
|
|
requests=searches,
|
|
|
|
|
|
fusion=Fusion(fusion=FusionProtocol.RRF)
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
if not fused_results.points:
|
|
|
|
|
|
debug("混合检索未找到任何文档")
|
|
|
|
|
|
return []
|
|
|
|
|
|
|
|
|
|
|
|
# 4. 收集 parent_id 和对应最高得分
|
|
|
|
|
|
parent_score_map = {}
|
|
|
|
|
|
parent_ids = set()
|
|
|
|
|
|
child_point_map = {} # 保存子文档点用于降级
|
|
|
|
|
|
|
|
|
|
|
|
for point in fused_results.points:
|
|
|
|
|
|
parent_id = point.payload.get("parent_id", point.id)
|
|
|
|
|
|
score = point.score
|
|
|
|
|
|
|
|
|
|
|
|
# 同一个 parent_id 只保留最高得分
|
|
|
|
|
|
if parent_id not in parent_score_map or score > parent_score_map[parent_id]:
|
|
|
|
|
|
parent_score_map[parent_id] = score
|
|
|
|
|
|
parent_ids.add(parent_id)
|
|
|
|
|
|
child_point_map[parent_id] = point
|
|
|
|
|
|
|
|
|
|
|
|
# 5. 批量查询父文档
|
|
|
|
|
|
# 首先尝试从 Qdrant 直接查询(因为父文档可能也存在 Qdrant 中)
|
|
|
|
|
|
parent_docs = []
|
|
|
|
|
|
found_parent_ids = set()
|
|
|
|
|
|
|
|
|
|
|
|
try:
|
|
|
|
|
|
parent_points = self.client.retrieve(
|
|
|
|
|
|
collection_name=self.collection_name,
|
|
|
|
|
|
ids=list(parent_ids),
|
|
|
|
|
|
with_payload=True
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
# 处理找到的父文档
|
|
|
|
|
|
for point in parent_points:
|
|
|
|
|
|
doc = Document(
|
|
|
|
|
|
page_content=point.payload.pop("text", ""),
|
|
|
|
|
|
metadata=point.payload
|
|
|
|
|
|
)
|
|
|
|
|
|
parent_docs.append(doc)
|
|
|
|
|
|
found_parent_ids.add(point.id)
|
|
|
|
|
|
|
|
|
|
|
|
except Exception as e:
|
|
|
|
|
|
warning(f"从 Qdrant 查询父文档失败: {e}")
|
|
|
|
|
|
|
|
|
|
|
|
# 6. 如果有 docstore,尝试从 docstore 查询剩余的父文档
|
|
|
|
|
|
if self.docstore and len(found_parent_ids) < len(parent_ids):
|
|
|
|
|
|
missing_parent_ids = parent_ids - found_parent_ids
|
|
|
|
|
|
try:
|
|
|
|
|
|
docstore_docs = self.docstore.mget(missing_parent_ids)
|
|
|
|
|
|
for doc_id, doc in zip(missing_parent_ids, docstore_docs):
|
|
|
|
|
|
if doc is not None:
|
|
|
|
|
|
parent_docs.append(doc)
|
|
|
|
|
|
found_parent_ids.add(doc_id)
|
|
|
|
|
|
except Exception as e:
|
|
|
|
|
|
warning(f"从 docstore 查询父文档失败: {e}")
|
|
|
|
|
|
|
|
|
|
|
|
# 7. 降级:对于仍未找到的父文档,用子文档本身代替
|
|
|
|
|
|
missing_parent_ids = parent_ids - found_parent_ids
|
|
|
|
|
|
if missing_parent_ids:
|
|
|
|
|
|
warning(f"以下 parent_id 未找到对应的父文档,将返回子文档本身: {missing_parent_ids}")
|
|
|
|
|
|
for parent_id in missing_parent_ids:
|
|
|
|
|
|
child_point = child_point_map.get(parent_id)
|
|
|
|
|
|
if child_point:
|
|
|
|
|
|
doc = Document(
|
|
|
|
|
|
page_content=child_point.payload.pop("text", ""),
|
|
|
|
|
|
metadata=child_point.payload
|
|
|
|
|
|
)
|
|
|
|
|
|
parent_docs.append(doc)
|
|
|
|
|
|
|
|
|
|
|
|
# 8. 按照得分降序排序,返回前 k 个
|
|
|
|
|
|
parent_docs_with_scores = [
|
|
|
|
|
|
(doc, parent_score_map.get(doc.metadata.get("id", doc.id), 0.0))
|
|
|
|
|
|
for doc in parent_docs
|
|
|
|
|
|
]
|
|
|
|
|
|
parent_docs_with_scores.sort(key=lambda x: x[1], reverse=True)
|
|
|
|
|
|
|
|
|
|
|
|
final_docs = [doc for doc, _ in parent_docs_with_scores[:self.search_k]]
|
|
|
|
|
|
debug(f"父子文档混合检索返回 {len(final_docs)} 个父文档")
|
|
|
|
|
|
|
|
|
|
|
|
return final_docs
|
|
|
|
|
|
|
|
|
|
|
|
async def _aget_relevant_documents(
|
|
|
|
|
|
self, query: str, *, run_manager: Optional[Any] = None
|
|
|
|
|
|
) -> List[Document]:
|
|
|
|
|
|
"""异步检索(当前调用同步版本)"""
|
|
|
|
|
|
return self._get_relevant_documents(query, run_manager=run_manager)
|
2026-04-21 11:02:16 +08:00
|
|
|
|
|
|
|
|
|
|
|
2026-05-04 02:01:22 +08:00
|
|
|
|
def create_hybrid_retriever(
|
2026-04-21 11:02:16 +08:00
|
|
|
|
collection_name: str,
|
2026-05-04 02:01:22 +08:00
|
|
|
|
search_k: int = DEFAULT_SEARCH_K,
|
|
|
|
|
|
embeddings: Optional[Embeddings] = None,
|
2026-04-21 11:02:16 +08:00
|
|
|
|
) -> BaseRetriever:
|
|
|
|
|
|
"""
|
2026-05-04 02:01:22 +08:00
|
|
|
|
创建混合检索器(稠密向量 + BM25 稀疏向量)。
|
|
|
|
|
|
|
|
|
|
|
|
这是默认推荐的检索方式,效果最优。
|
|
|
|
|
|
|
2026-04-21 11:02:16 +08:00
|
|
|
|
Args:
|
2026-04-29 10:52:01 +08:00
|
|
|
|
collection_name: Qdrant 集合名称
|
2026-05-04 02:01:22 +08:00
|
|
|
|
search_k: 检索返回结果数
|
|
|
|
|
|
embeddings: 可选的嵌入模型实例。若未提供,将自动获取统一嵌入服务。
|
|
|
|
|
|
|
2026-04-21 11:02:16 +08:00
|
|
|
|
Returns:
|
2026-05-04 02:01:22 +08:00
|
|
|
|
HybridRetriever 实例
|
2026-04-21 11:02:16 +08:00
|
|
|
|
"""
|
2026-05-04 02:01:22 +08:00
|
|
|
|
# 默认使用统一嵌入服务
|
2026-04-29 10:52:01 +08:00
|
|
|
|
if embeddings is None:
|
|
|
|
|
|
embeddings = get_embedding_service()
|
|
|
|
|
|
info("✅ 使用统一嵌入服务(本地 llama.cpp → 智谱 API 自动降级)")
|
2026-05-04 02:01:22 +08:00
|
|
|
|
|
|
|
|
|
|
# 创建向量存储
|
|
|
|
|
|
vector_store = QdrantVectorStore(collection_name=collection_name, embeddings=embeddings)
|
|
|
|
|
|
|
2026-04-29 10:52:01 +08:00
|
|
|
|
# 验证集合是否存在
|
2026-04-21 11:02:16 +08:00
|
|
|
|
try:
|
2026-05-04 02:01:22 +08:00
|
|
|
|
vector_store.get_client().get_collection(collection_name)
|
2026-04-21 11:02:16 +08:00
|
|
|
|
except UnexpectedResponse as e:
|
|
|
|
|
|
if e.status_code == 404:
|
2026-04-29 10:52:01 +08:00
|
|
|
|
warning(f"⚠️ Qdrant 集合 '{collection_name}' 不存在,请先创建并索引文档")
|
|
|
|
|
|
raise ValueError(f"Qdrant 集合 '{collection_name}' 不存在")
|
2026-04-21 11:02:16 +08:00
|
|
|
|
raise
|
2026-05-04 02:01:22 +08:00
|
|
|
|
|
|
|
|
|
|
info(f"✅ Qdrant 混合检索器初始化成功(search_k={search_k})")
|
|
|
|
|
|
return HybridRetriever(
|
2026-04-21 11:02:16 +08:00
|
|
|
|
collection_name=collection_name,
|
2026-05-04 02:01:22 +08:00
|
|
|
|
vector_store=vector_store,
|
|
|
|
|
|
search_k=search_k
|
2026-04-21 11:02:16 +08:00
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-05-04 02:01:22 +08:00
|
|
|
|
def create_parent_hybrid_retriever(
|
2026-04-21 11:02:16 +08:00
|
|
|
|
collection_name: str,
|
2026-05-04 02:01:22 +08:00
|
|
|
|
search_k: int = DEFAULT_PARENT_SEARCH_K,
|
|
|
|
|
|
embeddings: Optional[Embeddings] = None,
|
|
|
|
|
|
use_docstore: bool = True,
|
2026-04-21 11:02:16 +08:00
|
|
|
|
) -> BaseRetriever:
|
|
|
|
|
|
"""
|
2026-05-04 02:01:22 +08:00
|
|
|
|
创建父子文档混合检索器(默认推荐)。
|
|
|
|
|
|
|
|
|
|
|
|
检索流程:
|
|
|
|
|
|
1. 混合检索找到相关子文档
|
|
|
|
|
|
2. 根据 parent_id 找到对应的父文档
|
|
|
|
|
|
3. 去重并返回父文档
|
|
|
|
|
|
|
2026-04-21 11:02:16 +08:00
|
|
|
|
Args:
|
2026-05-04 02:01:22 +08:00
|
|
|
|
collection_name: Qdrant 集合名称
|
|
|
|
|
|
search_k: 最终返回的父文档数
|
|
|
|
|
|
embeddings: 可选的嵌入模型实例
|
|
|
|
|
|
use_docstore: 是否使用 PostgreSQL docstore 存储父文档
|
|
|
|
|
|
|
2026-04-21 11:02:16 +08:00
|
|
|
|
Returns:
|
2026-05-04 02:01:22 +08:00
|
|
|
|
ParentHybridRetriever 实例
|
2026-04-21 11:02:16 +08:00
|
|
|
|
"""
|
2026-05-04 02:01:22 +08:00
|
|
|
|
# 默认使用统一嵌入服务
|
2026-05-03 17:58:21 +08:00
|
|
|
|
if embeddings is None:
|
|
|
|
|
|
embeddings = get_embedding_service()
|
|
|
|
|
|
info("✅ 使用统一嵌入服务(本地 llama.cpp → 智谱 API 自动降级)")
|
2026-05-04 02:01:22 +08:00
|
|
|
|
|
|
|
|
|
|
# 创建向量存储
|
|
|
|
|
|
vector_store = QdrantVectorStore(collection_name=collection_name, embeddings=embeddings)
|
|
|
|
|
|
|
2026-05-03 17:58:21 +08:00
|
|
|
|
# 验证集合是否存在
|
2026-05-03 17:46:38 +08:00
|
|
|
|
try:
|
2026-05-04 02:01:22 +08:00
|
|
|
|
vector_store.get_client().get_collection(collection_name)
|
2026-05-03 17:58:21 +08:00
|
|
|
|
except UnexpectedResponse as e:
|
|
|
|
|
|
if e.status_code == 404:
|
|
|
|
|
|
warning(f"⚠️ Qdrant 集合 '{collection_name}' 不存在,请先创建并索引文档")
|
|
|
|
|
|
raise ValueError(f"Qdrant 集合 '{collection_name}' 不存在")
|
|
|
|
|
|
raise
|
2026-05-04 02:01:22 +08:00
|
|
|
|
|
|
|
|
|
|
# 创建 docstore(如果需要)
|
|
|
|
|
|
docstore = None
|
|
|
|
|
|
if use_docstore:
|
|
|
|
|
|
try:
|
|
|
|
|
|
docstore, _ = create_docstore()
|
|
|
|
|
|
info("✅ 文档存储初始化成功(PostgreSQL)")
|
|
|
|
|
|
except Exception as e:
|
|
|
|
|
|
warning(f"⚠️ 文档存储初始化失败,将不使用 docstore: {e}")
|
|
|
|
|
|
|
|
|
|
|
|
info(f"✅ Qdrant 父子文档混合检索器初始化成功(search_k={search_k})")
|
|
|
|
|
|
return ParentHybridRetriever(
|
2026-05-03 18:12:20 +08:00
|
|
|
|
collection_name=collection_name,
|
2026-05-04 02:01:22 +08:00
|
|
|
|
vector_store=vector_store,
|
|
|
|
|
|
search_k=search_k,
|
|
|
|
|
|
docstore=docstore
|
2026-05-03 17:58:21 +08:00
|
|
|
|
)
|
2026-05-03 17:46:38 +08:00
|
|
|
|
|
2026-05-03 18:12:20 +08:00
|
|
|
|
|
2026-05-04 02:01:22 +08:00
|
|
|
|
# 别名:默认就是父子文档混合检索
|
|
|
|
|
|
create_retriever = create_parent_hybrid_retriever
|