why100k/buildVectors2.py

# coding=utf-8
from langchain.document_loaders import UnstructuredFileLoader
from langchain.document_loaders import UnstructuredMarkdownLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import Chroma
from langchain.embeddings.huggingface import HuggingFaceEmbeddings
from langchain_community.document_loaders import PyPDFLoader
from langchain_community.document_loaders import MathpixPDFLoader
from langchain_community.document_loaders import UnstructuredPDFLoader
from langchain_community.document_loaders import AzureAIDocumentIntelligenceLoader
from langchain_community.document_loaders import Docx2txtLoader
from langchain_community.document_loaders.csv_loader import CSVLoader

from langchain.text_splitter import MarkdownHeaderTextSplitter
from tqdm import tqdm
from sentence_transformers import SentenceTransformer, util
import os
import chardet

import erniebot
import numpy as np
from langchain_embedding_ErnieBotSDK import ErnieEmbeddings

erniebot.api_type = "aistudio"
erniebot.access_token = "ff1531c8c0f429f92adbc2eaed2e23bfb5349e0f"

embeddings=ErnieEmbeddings(access_token="ff1531c8c0f429f92adbc2eaed2e23bfb5349e0f", chunk_size=1)

# 获取文件路径函数
def get_files(dir_path):
    # args：dir_path，目标文件夹路径
    file_list = []
    for filepath, dirnames, filenames in os.walk(dir_path):
        # os.walk 函数将递归遍历指定文件夹
        for filename in filenames:
            # 通过后缀名判断文件类型是否满足要求
            if filename.endswith(".md"):
                # 如果满足要求，将其绝对路径加入到结果列表
                file_list.append(os.path.join(filepath, filename))
            elif filename.endswith(".txt"):
                file_list.append(os.path.join(filepath, filename))
            #elif filename.endswith(".pdf"):
                #file_list.append(os.path.join(filepath, filename))
            elif filename.endswith(".docx"):
                file_list.append(os.path.join(filepath, filename))
            elif filename.endswith(".csv"):
                file_list.append(os.path.join(filepath, filename))

    return file_list


# 加载文件函数
def get_text(dir_path):
    file_lst = get_files(dir_path)
    docs = []
    for one_file in tqdm(file_lst):
        file_type = one_file.split('.')[-1]

        # 尝试检测文件编码
        with open(one_file, 'rb') as f:
            rawdata = f.read()
        encoding = chardet.detect(rawdata)['encoding']
        print(f"Detected encoding for {one_file}: {encoding}")

        # 根据文件类型创建适当的加载器，并指定编码
        if file_type == 'md':
            loader = UnstructuredMarkdownLoader(one_file, encoding=encoding)
        elif file_type == 'txt':
            loader = UnstructuredFileLoader(one_file, encoding=encoding)
        #elif file_type == 'pdf':
            # loader = PyPDFLoader(one_file)
            # loader = MathpixPDFLoader(one_file)
            #loader = UnstructuredPDFLoader(one_file)
        elif file_type == 'docx':
            loader = Docx2txtLoader(one_file, encoding=encoding)
        elif file_type == 'csv':
            loader = CSVLoader(one_file, encoding=encoding)
        else:
            continue

        # 加载文档
        try:
            docs.extend(loader.load())
        except UnicodeDecodeError as e:
            print(f"Failed to load {one_file} due to encoding error: {str(e)}")
            continue

    return docs


# 目标文件夹
tar_dir = [
    "files/",
]

# 加载目标文件
docs = []
for dir_path in tar_dir:
    print(get_text(dir_path))
    docs.extend(get_text(dir_path))

print(docs)

#embeddings = HuggingFaceEmbeddings(model_name="/mnt/sdb/zhaoyuan/rongrunxiang/acge_text_embedding")

# 构建向量数据库
# 定义持久化路径
persist_directory = 'data_base/vector_db/chroma'
# 加载数据库
vectordb = Chroma.from_documents(
    documents=docs,
    embedding=embeddings,
    persist_directory=persist_directory  # 允许我们将persist_directory目录保存到磁盘上
)
# 将加载的向量数据库持久化到磁盘上
vectordb.persist()