import os
# 导入 Document Loadersfrom langchain_community.document_loaders import PyPDFLoader
# Load Pdfbase_dir = '.\\easysearch' # 文档的存放目录docs = []for file in os.listdir(base_dir):    file_path = os.path.join(base_dir, file)    if file.endswith('.pdf'):        loader = PyPDFLoader(file_path)        documents.extend(loader.load())

# 2.将Documents切分成块from langchain.text_splitter import RecursiveCharacterTextSplittertext_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=20)chunked_documents = text_splitter.split_documents(docs)

# 3. 定义embedding模型from langchain_community.embeddings import OllamaEmbeddingsollama_emb = OllamaEmbeddings(    model="mxbai-embed-large",)
# 4. 定义 easysearch 集群的信息，以及存放向量的索引名称 infinifrom langchain_community.vectorstores import EcloudESVectorStoreES_URL = "https://192.168.56.3:9200"USER = "admin"PASSWORD = "e5ac1b537785ae27c187"indexname = "infini"
docsearch = EcloudESVectorStore.from_documents(    chunked_documents,    ollama_emb,    es_url=ES_URL,    user=USER,    password=PASSWORD,    index_name=indexname,    verify_certs=False,)

# 5. Retrieval 准备模型和Retrieval链import logging# MultiQueryRetriever工具from langchain.retrievers.multi_query import MultiQueryRetriever# RetrievalQA链from langchain.chains import RetrievalQA
# # 设置Logginglogging.basicConfig()logging.getLogger('langchain.retrievers.multi_query').setLevel(logging.INFO)
# # 实例化一个大模型工具from langchain_community.chat_models import ChatOllamallm = ChatOllama(model="qwen2:latest")
from langchain.prompts import PromptTemplatemy_template = PromptTemplate(    input_variables=["question"],    template="""You are an AI language model assistant. Your task is    to generate 3 different versions of the given user    question in Chinese to retrieve relevant documents from a vector  database.    By generating multiple perspectives on the user question,    your goal is to help the user overcome some of the limitations    of distance-based similarity search. Provide these alternative    questions separated by newlines. Original question: {question}""",)
# # 实例化一个MultiQueryRetrieverretriever_from_llm = MultiQueryRetriever.from_llm(retriever=docsearch.as_retriever(), llm=llm,prompt=my_template,include_original=True)
# # 实例化一个RetrievalQA链qa_chain = RetrievalQA.from_chain_type(llm,retriever=retriever_from_llm)

# 6. Q&A系统的UI实现from flask import Flask, request, render_templateapp = Flask(__name__) # Flask APP
@app.route('/', methods=['GET', 'POST'])def home():    if request.method == 'POST':
        # 接收用户输入作为问题        question = request.form.get('question')
        # RetrievalQA链 - 读入问题，生成答案        result = qa_chain({"query": question})
        # 把大模型的回答结果返回网页进行渲染        return render_template('index.html', result=result)
    return render_template('index.html')
if __name__ == "__main__":    app.run(host='0.0.0.0',debug=True,port=5000)