app.py

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

api_url="http://mail.digt.ru:7001/v1/chat/completions"
api_key="MyDiscord"

def download_pdf(url, output_path):
    urllib.request.urlretrieve(url, output_path)


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


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'[Страница {idx+start_page}]' + ' ' + '"' + chunk + '"'
            chunks.append(chunk)
    return chunks

class SemanticSearch:
    
    def __init__(self):
       #self.use = hub.load('https://tfhub.dev/google/universal-sentence-encoder/4')
       self.use = hub.load('https://tfhub.dev/google/universal-sentence-encoder-multilingual-large/3')
       #self.use = hub.load('/app/models/')
       self.fitted = False
    
    
    def fit(self, data, batch=1000, n_neighbors=2):
        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
    
    
    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
    
    
    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



def load_recommender(path, start_page=1):
    global recommender
    texts = pdf_to_text(path, start_page=start_page)
    chunks = text_to_chunks(texts, start_page=start_page)
    recommender.fit(chunks)
    return 'Corpus Loaded.'

def generate_text(prompt):

    data = {
        "frequency_penalty": 0,
        "model": "gpt-3.5-turbo",
        "presence_penalty": 0,
        "temperature":1,
        "top_p": 1,
        "messages":[{"role":"system","content":"You are ChatGPT, a large language model trained by OpenAI.\nCarefully heed the user's instructions. \nRespond using Markdown."},{"role":"user", "content": prompt}]
        }
    
    #print(data)
    r = requests.post(api_url, headers = {'Authorization': 'Bearer ' + api_key}, json = data) 
    completions = r.json()
    message = completions.get("choices")[0].get("message").get("content")
    print(message)
    return message

def generate_answer(question):
    topn_chunks = recommender(question)
    prompt = ""
    prompt += 'Результаты поиска:\n\n'
    for c in topn_chunks:
        prompt += c + '\n\n'
        
    prompt += (
        "Инструкция: Составь исчерпывающий ответ на вопрос, используя приведенные результаты поиска. "
        "Включай только информацию, найденную в результатах поиска, и "
        "не добавляй никакой дополнительной информации. Убедись, что ответ правильный, и не выводи ложный контент. "
        "Если текст не относится к воросу, просто укажи 'Информация не найдена в этом документе'. Игнорируй "
        "результаты поиска, которые не имеют никакого отношения к вопросу. Отвечайте только на то, о чем тебя спрашивают. "
        "Ответ должен быть коротким и лаконичным. Отвечай шаг за шагом.\n\n"
    )

    prompt += f"Запрос: {question}\nОтвет:"
    print('prompt','->', prompt)
    answer = generate_text(prompt)
    return answer

class Empty():
    pass

def question_answer(file, question):
    if file == None:
        load_recommender('/app/corpus.pdf')
    else:
        old_file_name = file.name
        file_name = file.name
        file_name = file_name[:-12] + file_name[-4:]
        os.rename(old_file_name, file_name)
        load_recommender(file_name)

    if question.strip() == '':
        return '[ERROR]: Question field is empty'

    return generate_answer(question)


recommender = SemanticSearch()

title = 'PDF GPT'
description = """ PDF GPT позволяет вам общаться в чате с вашим PDF-файлом, используя Universal Sentence Encoder и Open AI"""

with gr.Blocks() as demo:

    gr.Markdown(f'<center><h1>{title}</h1></center>')
    gr.Markdown(description)

    with gr.Row():
        
        with gr.Group():
            gr.Markdown("<center><h4><h4></center>")
            file = gr.File(label='Загрузите ваш PDF', file_types=['.pdf'])
            gr.Markdown("<center><h4><h4></center>")
            question = gr.Textbox(label='Введите ваш вопрос')
            btn = gr.Button(value='Submit')
            btn.style(full_width=True)

        with gr.Group():
            answer = gr.Textbox(label='Ответ на ваш вопрос :')

        btn.click(question_answer, inputs=[file, question], outputs=[answer])
demo.launch()