修改rag，实现混合检索

tools/test/check_qdrant.py

@@ -0,0 +1,80 @@
+#!/usr/bin/env python3
+"""
+检查 Qdrant 集合里的数据结构
+"""
+
+import asyncio
+import os
+import sys
+
+# 添加项目根目录到 Python 路径
+project_root = os.path.join(os.path.dirname(__file__), "..", "..")
+sys.path.insert(0, os.path.join(project_root, "backend"))
+sys.path.insert(0, project_root)
+
+from rag_core import QdrantVectorStore
+from app.model_services import get_embedding_service
+
+
+def check_qdrant_data():
+    """检查 Qdrant 中的数据结构"""
+    print("="*70)
+    print("检查 Qdrant 中的数据结构...")
+    print("="*70)
+    
+    embeddings = get_embedding_service()
+    vs = QdrantVectorStore(collection_name="rag_documents", embeddings=embeddings)
+    client = vs.get_qdrant_client()
+    
+    # 先获取几个点看看 payload 结构
+    print("\n获取 5 个随机文档:")
+    results = client.scroll(
+        collection_name="rag_documents",
+        limit=5,
+        with_payload=True,
+        with_vectors=True
+    )
+    
+    for i, point in enumerate(results[0], 1):
+        print(f"\n{i}. ID: {point.id}")
+        print(f"   Payload: {point.payload}")
+        print(f"   Payload 键: {list(point.payload.keys())}")
+        if "text" in point.payload:
+            text = point.payload["text"]
+            print(f"   Text 长度: {len(text)}")
+            print(f"   Text 预览: {text[:150]}...")
+        if "page_content" in point.payload:
+            print(f"   page_content: {point.payload['page_content'][:150]}...")
+        
+        # 看看向量
+        if point.vector:
+            print(f"   向量存在: {type(point.vector)}")
+            if isinstance(point.vector, dict):
+                print(f"   向量键: {list(point.vector.keys())}")
+
+
+def check_sparse_embedder():
+    """检查稀疏嵌入器"""
+    from rag_core import get_sparse_embedder
+    
+    print("\n" + "="*70)
+    print("检查稀疏嵌入器...")
+    print("="*70)
+    
+    sparse_embedder = get_sparse_embedder()
+    
+    print(f"\n稀疏嵌入器: {sparse_embedder}")
+    print(f"Vocabulary 大小: {len(sparse_embedder.model.vocab)}")
+    print(f"示例查询: '冬天 食物'")
+    
+    # 用中文试试
+    sparse_vec = sparse_embedder.embed_query("冬天 食物")
+    print(f"\n生成的稀疏向量:")
+    print(f"  索引数量: {len(sparse_vec['indices'])}")
+    print(f"  索引: {sparse_vec['indices'][:10]}")
+    print(f"  值: {sparse_vec['values'][:10]}")
+
+
+if __name__ == "__main__":
+    check_qdrant_data()
+    check_sparse_embedder()