145 lines
5.3 KiB
Python
145 lines
5.3 KiB
Python
import warnings
|
||
import os
|
||
import time
|
||
from Config import Config
|
||
from elasticsearch import Elasticsearch
|
||
from langchain_core.documents import Document
|
||
from Util.VectorUtil import text_to_vector_db # 导入向量化工具函数
|
||
from langchain_openai import OpenAIEmbeddings # 直接导入嵌入模型
|
||
from pydantic import SecretStr # 用于包装API密钥
|
||
|
||
# 抑制HTTPS相关警告
|
||
warnings.filterwarnings('ignore', message='Connecting to .* using TLS with verify_certs=False is insecure')
|
||
warnings.filterwarnings('ignore', message='Unverified HTTPS request is being made to host')
|
||
|
||
|
||
def insert_long_text_to_es(long_text: str, tags: list = None) -> bool:
|
||
"""
|
||
将长文本向量化并插入到Elasticsearch
|
||
|
||
参数:
|
||
long_text: 要插入的长文本
|
||
tags: 可选的标签列表
|
||
|
||
返回:
|
||
bool: 插入是否成功
|
||
"""
|
||
try:
|
||
# 1. 初始化Elasticsearch连接
|
||
es = Elasticsearch(
|
||
hosts=Config.ES_CONFIG['hosts'],
|
||
basic_auth=Config.ES_CONFIG['basic_auth'],
|
||
verify_certs=False
|
||
)
|
||
|
||
# 2. 检查索引是否存在，不存在则创建
|
||
index_name = Config.ES_CONFIG['index_name']
|
||
if not es.indices.exists(index=index_name):
|
||
# 定义mapping结构
|
||
mapping = {
|
||
"mappings": {
|
||
"properties": {
|
||
"embedding": {
|
||
"type": "dense_vector",
|
||
"dims": 1024, # 根据实际embedding维度调整
|
||
"index": True,
|
||
"similarity": "l2_norm"
|
||
},
|
||
"user_input": {"type": "text"},
|
||
"tags": {
|
||
"type": "object",
|
||
"properties": {
|
||
"tags": {"type": "keyword"},
|
||
"full_content": {"type": "text"}
|
||
}
|
||
},
|
||
"timestamp": {"type": "date"}
|
||
}
|
||
}
|
||
}
|
||
es.indices.create(index=index_name, body=mapping)
|
||
print(f"索引 '{index_name}' 创建成功")
|
||
|
||
# 3. 创建文档对象
|
||
docs = [Document(page_content=long_text, metadata={"source": "user_provided_text"})]
|
||
|
||
# 4. 获取当前时间
|
||
timestamp = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
|
||
|
||
# 5. 准备标签
|
||
if tags is None:
|
||
tags = ["general_text"]
|
||
tags_dict = {"tags": tags, "full_content": long_text}
|
||
|
||
# 6. 直接创建嵌入模型并生成向量
|
||
embeddings = OpenAIEmbeddings(
|
||
model=Config.EMBED_MODEL_NAME,
|
||
base_url=Config.EMBED_BASE_URL,
|
||
api_key=SecretStr(Config.EMBED_API_KEY)
|
||
)
|
||
|
||
# 7. 生成文本嵌入向量
|
||
embedding = embeddings.embed_documents([long_text])[0]
|
||
|
||
# 8. 插入数据到Elasticsearch
|
||
doc = {
|
||
'tags': tags_dict,
|
||
'user_input': long_text[:500], # 取前500个字符作为摘要
|
||
'timestamp': timestamp,
|
||
'embedding': embedding
|
||
}
|
||
es.index(index=index_name, document=doc)
|
||
print(f"长文本数据插入成功")
|
||
return True
|
||
except Exception as e:
|
||
print(f"插入数据失败: {e}")
|
||
return False
|
||
|
||
|
||
def process_text_directory(txt_dir: str) -> None:
|
||
"""
|
||
处理指定目录下的所有文本文件，将其向量化并插入到Elasticsearch
|
||
|
||
参数:
|
||
txt_dir: 包含文本文件的目录路径
|
||
"""
|
||
for filename in os.listdir(txt_dir):
|
||
if filename.endswith('.txt'):
|
||
filepath = os.path.join(txt_dir, filename)
|
||
with open(filepath, 'r', encoding='utf-8') as f:
|
||
full_content = f.read()
|
||
|
||
if not full_content:
|
||
print(f"跳过空文件: {filename}")
|
||
continue
|
||
|
||
print(f"正在处理文件: {filename}")
|
||
|
||
# 提取标签
|
||
x = filename.split("_")
|
||
if len(x) >= 2:
|
||
selected_tags = [x[0] + "_" + x[1]]
|
||
else:
|
||
selected_tags = ["uncategorized"]
|
||
|
||
# 插入文本
|
||
insert_long_text_to_es(full_content, selected_tags)
|
||
|
||
|
||
def main():
|
||
# 示例1：插入单个长文本
|
||
long_text = """混凝土是一种广泛使用的建筑材料，由水泥、砂、石子和水混合而成。它具有高强度、耐久性和良好的可塑性，被广泛应用于建筑、桥梁、道路等土木工程领域。
|
||
|
||
混凝土的历史可以追溯到古罗马时期，当时人们使用火山灰、石灰和碎石混合制成类似混凝土的材料。现代混凝土技术始于19世纪，随着波特兰水泥的发明而得到快速发展。
|
||
|
||
混凝土的性能取决于其配合比，包括水灰比、砂率等参数。水灰比是影响混凝土强度的关键因素，较小的水灰比通常会产生更高强度的混凝土。"""
|
||
insert_long_text_to_es(long_text, tags=["construction", "materials"])
|
||
|
||
# 示例2：处理目录中的所有文本文件
|
||
# txt_dir = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), 'Txt')
|
||
# process_text_directory(txt_dir)
|
||
|
||
|
||
if __name__ == "__main__":
|
||
main()
|