极客时间0期AI大模型应用开发实战营作业参考_ai 大模型应用开发实战营

极客时间《AI 大模型应用开发实战营》

最近在极客时间学习《AI 大模型应用开发实战营》，通过输入提示词让chatgpt完成了简单的pdf上传、下载的前端页面.目前上传英文pdf文档后，服务器后台会解析pdf文档，并翻译成中文pdf文档，可以下载下来。

prompt提示词内容如下：‘帮我生成上传、下载pdf文件的前端页面，表单请求触发，后端上传路由是localhost:5000/upload，下载路由是localhost:5000/download/<filename>,上传文件和下载文件文件名是相同的’。页面截图如下：

 web接口开发采用的是flask框架，接口开发代码，也是chatgpt自动生成的。提示词如下：’请用flask框架 写上传pdf和下载pdf的接口文件’。chatgpt自动生成的代码还需要修改，我在它生成的代码文件中添加ip和port的设置，并将翻译功能的主要代码添加进upload_file函数接口中，最终形成的代码文件如下：

from flask import Flask, request, send_file
import sys
import os

sys.path.append(os.path.dirname(os.path.abspath(__file__)))

from utils import ArgumentParser, ConfigLoader, LOG
from model import GLMModel, OpenAIModel
from translator import PDFTranslator
app = Flask(__name__)

# 设置上传文件存储目录
UPLOAD_FOLDER = 'uploads'
app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER

# 确保上传文件存储目录存在
if not os.path.exists(UPLOAD_FOLDER):
    os.makedirs(UPLOAD_FOLDER)


@app.route('/upload', methods=['POST'])
def upload_file():
    if 'file' not in request.files:
        return 'No file part'

    file = request.files['file']

    if file.filename == '':
        return 'No selected file'

    if file:
        filename = os.path.join(app.config['UPLOAD_FOLDER'], file.filename)
        file.save(filename)
        argument_parser = ArgumentParser()
        args = argument_parser.parse_arguments()
        config_loader = ConfigLoader(args.config)

        config = config_loader.load_config()

        model_name = args.openai_model if args.openai_model else config['OpenAIModel']['model']
        api_key = args.openai_api_key if args.openai_api_key else config['OpenAIModel']['api_key']
        model = OpenAIModel(model=model_name, api_key=api_key)

        pdf_file_path = filename
        file_format = args.file_format if args.file_format else config['common']['file_format']

        # 实例化 PDFTranslator 类，并调用 translate_pdf() 方法
        translator = PDFTranslator(model)
        translator.translate_pdf(pdf_file_path, file_format)
        return 'File uploaded successfully'


@app.route('/download/<filename>', methods=['GET'])
def download_file(filename):
    file_path = os.path.join(app.config['UPLOAD_FOLDER'], filename)

    if os.path.exists(file_path):
        return send_file(file_path, as_attachment=True)
    else:
        return 'File not found'


if __name__ == '__main__':
    ip='127.0.0.1'
    post=5000
    app.run(ip=ip,post=post)

在pdf的解析函数中，我先把一个page文件中的所有表格提取出来，使用pdfplumber包提供的crop方法将表格和文字全部剪切成各自部分的文件，并存入切片数组中，再从切片数组中提取各部分文件，最后按照文件类型提取文字和表格存入page类中。pdf解析的代码函数如下：

def parse_pdf(self, pdf_file_path: str, pages: Optional[int] = None) -> Book:
    book = Book(pdf_file_path)

    with pdfplumber.open(pdf_file_path) as pdf:
        if pages is not None and pages > len(pdf.pages):
            raise PageOutOfRangeException(len(pdf.pages), pages)

        if pages is None:
            pages_to_parse = pdf.pages
        else:
            pages_to_parse = pdf.pages[:pages]

        for pdf_page in pages_to_parse:
            page = Page()

            # Store the original text content
            tables=pdf_page.find_tables(table_settings={})

            # Remove each cell's content from the original text
            content = []

            for i in range(len(tables) - 1):
                table_page = pdf_page.crop(tables[i].bbox, relative=False, strict=True)
                content.append(('table',table_page.extract_table()))
                middle_page = pdf_page.crop((0, tables[i].bbox[-1], pdf_page.width, tables[i + 1].bbox[1]), relative=False,
                                        strict=True)
                content.append(('text',middle_page.extract_text()))
            if len(tables) > 0:
                init_page = pdf_page.crop((0, 0, pdf_page.width, tables[0].bbox[1]), relative=False, strict=True)
                init_text = init_page.extract_text()
                content.insert(0, ('text',init_text))
                last_table_page = pdf_page.crop(tables[-1].bbox, relative=False, strict=True)
                content.append(('table',last_table_page.extract_table()))
                last_page = pdf_page.crop((0, tables[-1].bbox[-1], pdf_page.width, pdf_page.height), relative=False,
                                      strict=True)
                content.append(('text',last_page.extract_text()))
            else:
                content.append(('text',pdf_page.extract_text()))

            # Handling text
            for item_type,item in content:
                # Remove empty lines and leading/trailing whitespaces
                if item_type=='text' and item:
                    raw_text_lines = item.splitlines()
                    cleaned_raw_text_lines = [line.strip() for line in raw_text_lines if line.strip()]
                    cleaned_raw_text = "\n".join(cleaned_raw_text_lines)
                    text_content = Content(content_type=ContentType.TEXT, original=cleaned_raw_text)
                    page.add_content(text_content)
                    LOG.debug(f"[raw_text]\n {cleaned_raw_text}")
                if item_type == 'table' and item:
                    table = TableContent(tables)
                    page.add_content(table)
                    LOG.debug(f"[table]\n{table}")
            book.add_page(page)

    return book