app.py

import re
import numpy as np
import tensorflow_hub as hub
import openai
import os
import tensorflow_text
from sklearn.neighbors import NearestNeighbors
import gradio as gr
import requests
import json
import fitz

#这里填写调用openai需要的密钥
openai.api_key = '9481961416fa4c8e883047c5679cf971'
openai.api_base = 'https://demopro-oai-we2.openai.azure.com/' 
openai.api_type = 'azure'
openai.api_version = '2022-12-01'

#将嵌套的列表展平
def flatten(_2d_list):
    flat_list = []
    for element in _2d_list:
        if type(element) is list:
            for item in element:
                flat_list.append(item)
        else:
            flat_list.append(element)
    return flat_list


def preprocess(text):
    text = text.replace('\n', ' ')
    text = re.sub('\s+', ' ', text)
    return text

#将pdf文档按段落分
# def pdf_to_text(path):
#     doc = pdfplumber.open(path)
#     pages = doc.pages
#     text_list=[]
#     for page,d in enumerate(pages):
#         d=d.extract_text()
#         d=preprocess(d)
#         text_list.append(d)  
#     doc.close()   
  
#     return text_list



def pdf_to_text(path, start_page=1, end_page=None):
    doc = fitz.open(path)
    total_pages = doc.page_count

    if end_page is None:
        end_page = total_pages

    text_list = []

    for i in range(start_page - 1, end_page):
        text = doc.load_page(i).get_text("text")
        text = preprocess(text)
        text_list.append(text)

    doc.close()
    return text_list


def text_to_chunks(texts, word_length=150, start_page=1):
    text_toks = [t.split(' ') for t in texts]
    page_nums = []
    chunks = []

    for idx, words in enumerate(text_toks):
        for i in range(0, len(words), word_length):
            chunk = words[i : i + word_length]
            if (
                (i + word_length) > len(words)
                and (len(chunk) < word_length)
                and (len(text_toks) != (idx + 1))
            ):
                text_toks[idx + 1] = chunk + text_toks[idx + 1]
                continue
            chunk = ' '.join(chunk).strip()
            chunk = f'[Page no. {idx+start_page}]' + ' ' + '"' + chunk + '"'
            chunks.append(chunk)
    return chunks
    
history=pdf_to_text('The Elements of Statisitcal Learning.pdf',start_page=20)
history=text_to_chunks(history,start_page=1)


def encoder(text):
    embed=openai.Embedding.create(input=text, engine="text-embedding-ada-002")
    return embed.get('data')[0].get('embedding')
    
    
#定义语义搜索类
class SemanticSearch:
    
    def __init__(self):
        #类初始化，使用google公司的多语言语句编码，第一次运行时需要十几分钟的时间下载
        self.use =hub.load('https://tfhub.dev/google/universal-sentence-encoder-multilingual/3')
        self.fitted = False
    
    def get_text_embedding(self, texts, batch=1000):
        embeddings = []
        for i in range(0, len(texts), batch):
            text_batch = texts[i : (i + batch)]
            emb_batch = self.use(text_batch)
            embeddings.append(emb_batch)
        embeddings = np.vstack(embeddings)
        return embeddings

    
    
    #K近邻算法，找到与问题最相似的 k 个段落,这里的 k 即n_neighbors=10
    def fit(self, data, batch=1000, n_neighbors=5):
        self.data = data
        self.embeddings = self.get_text_embedding(data, batch=batch)
        n_neighbors = min(n_neighbors, len(self.embeddings))
        self.nn = NearestNeighbors(n_neighbors=n_neighbors)
        self.nn.fit(self.embeddings)
        self.fitted = True
    
    #定义了该方法后，实例就可以被当作函数调用,text参数即用户提出的问题，inp_emb为其转化成的向量
    def __call__(self, text, return_data=True):
        inp_emb = self.use([text])
        
        
        neighbors = self.nn.kneighbors(inp_emb, return_distance=False)[0]
        
        if return_data:
            return [self.data[i] for i in neighbors]
        else:
            return neighbors
    
    
#openai的api接口，engine参数为我们选择的大语言模型，prompt即提示词
def generate_text(prompt, engine="text-davinci-003"):
    completions = openai.Completion.create(
        engine=engine,
        prompt=prompt,
        max_tokens=512,
        n=1,
        stop=None,
        temperature=0.7,
    )
    message = completions.choices[0].text
    return message


def generate_answer(question):
    #匹配与问题最相近的n个段落，前面定义了n=10
    topn_chunks = recommender(question)
    prompt = ""
    prompt += 'search results:\n\n'
    
    #把匹配到的段落加进提示词
    for c in topn_chunks:
        prompt += c + '\n\n'
    
    #提示词
    prompt += '''
    Instructions: 如果搜索结果中找不到相关信息，只需要回答'未在该文档中找到相关信息'。
    如果找到了相关信息，请使用中文回答，回答尽量精确简洁。并在句子的末尾使用[七年级上册/七年级下册页码]符号引用每个参考文献（每个结果的开头都有这个编号)
    如果不确定答案是否正确，就仅给出相似段落的来源，不要回复错误的答案。
    \n\nQuery: {question}\nAnswer: 
    '''
    
    prompt += f"Query: {question}\nAnswer:"
    answer = generate_text(prompt,"text-davinci-003")
    return answer


recommender = SemanticSearch()
recommender.fit(history)


#以下为web客户端搭建，运行后产生客户端界面
def ask_api(question):
    
    if question.strip() == '':
        return '[ERROR]: 未输入问题'

    return generate_answer(question)

title = 'Chat With Statistical Learning'
description = """ 该机器人将以Trevor Hastie等人所著的The Elements of Statistical Learning Data Mining, Inference, and Prediction
(即我们上课所用的课本)为主题回答你的问题，如果所问问题与书的内容无关，将会返回"未在该文档中找到相关信息"
"""

with gr.Blocks() as demo:
    gr.Markdown(f'<center><h1>{title}</h1></center>')
    gr.Markdown(description)

    with gr.Row():
        with gr.Group():
            question = gr.Textbox(label='请输入你的问题')
            btn = gr.Button(value='提交')
            btn.style(full_width=True)

        with gr.Group():
            answer = gr.Textbox(label='回答:')

        btn.click(
            ask_api,
            inputs=[question],
            outputs=[answer]
        )

#参数share=True会产生一个公开网页，别人可以通过访问该网页使用你的模型，前提是你需要正在运行这段代码（将自己的电脑当作服务器）
demo.launch()