'commit'

dsRag/Milvus/X1_create_collection.py

@@ -26,11 +26,11 @@ if utility.has_collection(collection_name):
 
 # 5. 定义集合的字段和模式
 fields = [
-    FieldSchema(name="id", dtype=DataType.INT64, is_primary=True, auto_id=True),  # 主键字段，自动生成 ID
+    FieldSchema(name="id", dtype=DataType.INT64, is_primary=True, auto_id=True),
-    FieldSchema(name="document_id", dtype=DataType.VARCHAR, max_length=64),  # 文档 ID
+    FieldSchema(name="tags", dtype=DataType.JSON),  # 改为JSON类型存储多个标签
-    FieldSchema(name="user_input", dtype=DataType.VARCHAR, max_length=65535),  # 用户问题
+    FieldSchema(name="user_input", dtype=DataType.VARCHAR, max_length=65535),
-    FieldSchema(name="timestamp", dtype=DataType.VARCHAR, max_length=32),  # 时间
+    FieldSchema(name="timestamp", dtype=DataType.VARCHAR, max_length=32),
-    FieldSchema(name="embedding", dtype=DataType.FLOAT_VECTOR, dim=MS_DIMENSION)  # 向量字段，维度为 200
+    FieldSchema(name="embedding", dtype=DataType.FLOAT_VECTOR, dim=MS_DIMENSION)
 ]
 schema_description = "Chat records collection with document_id , user_input, and timestamp"
 

dsRag/Milvus/X4_InsertMathData.py

@@ -46,14 +46,14 @@ for filename in os.listdir(txt_dir):
 
             # 5. 获取当前时间和会话ID
             timestamp = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
-            document_id = "MATH_DATA_1" # 史校长的这本书定义为 MATH_DATA_1
+            tags = {"tags": ["MATH_DATA_1", "小学数学"]} # 直接使用Python字典，Milvus会自动转换为JSON
 
             # 6. 将文本转换为嵌入向量
             embedding = text_to_embedding(content)
 
             # 7. 插入数据
             entities = [
-                [document_id],  # document_id
+                [tags],  # tags
                 [content],  # user_input
                 [timestamp],  # timestamp
                 [embedding]  # embedding

dsRag/Milvus/X5_select_all_data.py

@@ -21,7 +21,7 @@ try:
     # 使用 Milvus 的 query 方法查询所有数据
     results = collection_manager.collection.query(
         expr="",  # 空表达式表示查询所有数据
-        output_fields=["id", "document_id", "user_input",  "timestamp", "embedding"],  # 指定返回的字段
+        output_fields=["id", "tags", "user_input",  "timestamp", "embedding"],  # 指定返回的字段
         limit=1000  # 设置最大返回记录数
     )
     print("查询结果：")
@@ -29,13 +29,13 @@ try:
         for result in results:
             try:
                 # 获取字段值
-                document_id = result["document_id"]
+                tags = result["tags"]
                 user_input = result["user_input"]
                 timestamp = result["timestamp"]
                 embedding = result["embedding"]
                 # 打印结果
                 print(f"ID: {result['id']}")
-                print(f"文档 ID: {document_id}")
+                print(f"标签: {tags}")
                 print(f"用户问题: {user_input}")
                 print(f"时间: {timestamp}")
                 print(f"向量: {embedding[:5]}...")  # 只打印前 5 维向量

dsRag/Milvus/X6_search_near_data.py

@@ -55,7 +55,7 @@ search_params = {
 # 哪些文档查询，哪些不查询，我说了算！
 # 这样的话，我就可以打多个标签了！
 # expr = "document_id in ['MATH_DATA_1', 'MATH_DATA_2']"
-expr = "document_id in ['MATH_DATA_1']"
+expr = "tags['tags'] == 'MATH_DATA_1'"
 results = collection_manager.search(
     current_embedding,
     search_params,
@@ -74,7 +74,7 @@ if results:
                 # 查询非向量字段
                 record = collection_manager.query_by_id(hit.id)
                 print(f"ID: {hit.id}")
-                print(f"文档 ID: {record['document_id']}")
+                print(f"标签: {record['tags']}")
                 print(f"用户问题: {record['user_input']}")
                 print(f"时间: {record['timestamp']}")
                 print(f"距离: {hit.distance}")