| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138 |
- #!/usr/bin/env python3
- """
- 测试修复后的 Milvus 向量实现
- """
- import sys
- import os
- import json
- # 添加项目根目录到路径
- # 解决模块导入问题:添加项目根目录到 Python 路径
- current_script_path = os.path.abspath(__file__)
- script_dir = os.path.dirname(current_script_path)
- project_root = os.path.abspath(os.path.join(script_dir, "../../"))
- if project_root not in sys.path:
- sys.path.insert(0, project_root)
- def load_documents_from_file(json_path: str):
- """
- 从单个 JSON 文件读取所有 chunks,生成:
- [{'content': str, 'metadata': dict}, ...]
- 只保留 chunk 自身的 content 和 metadata,不混入文件级元数据
- """
- documents = []
- if not os.path.isfile(json_path):
- print(f"[WARN] JSON 文件不存在: {json_path}")
- return documents
- try:
- with open(json_path, "r", encoding="utf-8") as f:
- data = json.load(f)
- except Exception as e:
- print(f"[ERROR] 读取 JSON 文件失败: {json_path}, error: {e}")
- return documents
- chunks = data.get("chunks", [])
- if not isinstance(chunks, list):
- print(f"[WARN] 文件 {json_path} 中的 chunks 字段不是 list,跳过")
- return documents
- for idx, chunk in enumerate(chunks):
- if not isinstance(chunk, dict):
- continue
- # 提取 content
- content = chunk.get("content", "")
- if not content or not str(content).strip():
- # 空内容不入库
- continue
- # 只用 chunk 自己的 metadata
- metadata = chunk.get("metadata", {}) or {}
- # 如果你也想保留 chunk 的索引,可以打开下面这行
- # metadata["chunk_index"] = idx
- documents.append({
- "content": content,
- "metadata": metadata,
- })
- print(f"[INFO] 文件 {os.path.basename(json_path)} 提取出 {len(documents)} 条 chunk 文档")
- return documents
- def test_basic_functionality():
- """测试基本功能"""
- try:
- # 导入并初始化 MilvusVectorManager
- from foundation.database.base.vector.milvus_vector import MilvusVectorManager
- print("成功导入 MilvusVectorManager")
- # 初始化管理器
- manager = MilvusVectorManager()
- print("MilvusVectorManager 初始化成功")
- # 测试 text_to_vector 方法
- test_text = "桥梁建设技术"
- vector = manager.text_to_vector(test_text)
- print(f"text_to_vector 测试成功,向量维度: {len(vector)}")
- # ====== 关键改动:从文件夹读取所有 JSON 文件,生成 documents ======
- # 指定你的 JSON 文件夹路径
- json_dir = "data_pipeline/test_rawdata"
- all_documents = []
- if not os.path.isdir(json_dir):
- print(f"[ERROR] 目录不存在: {json_dir}")
- return False
- # 遍历文件夹下所有文件
- for filename in os.listdir(json_dir):
- # 只处理 .json 文件(如果你是其它后缀,改这里)
- if not filename.lower().endswith(".json"):
- continue
- json_path = os.path.join(json_dir, filename)
- docs = load_documents_from_file(json_path)
- if docs:
- all_documents.extend(docs)
- print(f"[INFO] 总共从目录 {json_dir} 中解析出 {len(all_documents)} 条文档")
- if not all_documents:
- print("[ERROR] 未从任何文件中解析到文档,停止测试")
- return False
- # ====== 关键改动结束 ======
- collection_name = "first_bfp_collection_test"
- print(f"\n测试 create_hybrid_collection 方法...")
- vectorstore = manager.create_hybrid_collection(
- collection_name=collection_name,
- documents=all_documents # ← 用目录里解析出的所有 documents
- )
- print("create_hybrid_collection 执行成功!")
- print(f"返回的 vectorstore 类型: {type(vectorstore)}")
- return True
- except Exception as e:
- print(f"测试失败: {e}")
- import traceback
- traceback.print_exc()
- return False
- if __name__ == "__main__":
- success = test_basic_functionality()
- print("\n" + "=" * 50)
- print(f"测试结果: {'成功' if success else '失败'}")
- if success:
- print("修复验证成功!")
- print("- text_to_vector 方法正常工作")
- print("- create_hybrid_collection 方法正常工作")
- print("- hybrid_search 方法正常工作")
|