ai_qback/deepseek/get_article3.py

# -*- coding: utf-8 -*-

import re
import json
from deepseek.ds_api import DS

from tools.new_mysql import MySQLUploader
from tools.loglog import logger, log_err_e
from tools.thread_pool_manager import pool_executor
from common.common_data import all_exchange_words
from common.split_text import split_text_to_word

from pydantic import BaseModel
from cachetools import TTLCache
from concurrent.futures import wait
from random import randint, shuffle
import json
import requests


def get_article_difficulty(article) -> int:
    """获取文章的难度值"""
    url = "http://qbank.yunzhixue.cn/api/article/analysis"

    data = {"body": article, "question": ""}
    try:
        response = requests.post(url, json=data)
    except Exception as e:
        log_err_e(e, msg="获取文章难度值；")
        return 0

    if response.status_code == 200:
        difficult_value = response.json()['data']['difficult']
        return difficult_value
    else:
        logger.error(f"错误状态码{response.status_code}")


def find_interval(number):
    """
    判断一个数字属于哪个难度等级区间。31级是例外情况，需要排查

    :param number: 要检查的数字。
    :return: 返回包含该数字的区间，如果没有找到，则返回 None。
    """
    intervals = [(1, 200), (201, 250), (251, 300), (301, 350), (351, 400), (401, 450), (451, 550), (551, 650), (651, 750), (751, 850),
                 (851, 950),
                 (951, 1100),
                 (1101, 1250), (1251, 1400), (1401, 1550), (1551, 1700), (1701, 1900), (1901, 2100), (2101, 2300), (2301, 2600),
                 (2601, 2900),
                 (2901, 3200),
                 (3201, 3500), (3501, 3900), (3901, 4300), (4301, 4700), (4701, 5100), (5101, 5500), (5501, 5900), (5901, 6500),
                 (6501, 99999)]
    for index, (start, end) in enumerate(intervals, start=1):
        if start <= number <= end:
            return index
    logger.error(f"文章难度判断不对:{number}")
    return 0


def parse_question(question_block):
    question_info = {}

    question_match = re.search(r'问题:\s*(.*)', question_block)
    if question_match:
        question_info['trunk'] = question_match.group(1).strip()

    analysis_match = re.search(r'解析:\s*(.*)', question_block)
    if analysis_match:
        question_info['analysis'] = analysis_match.group(1).strip()

    options_match = re.search(r'选项：(.*)', question_block)
    if options_match:
        options_text = options_match.group(1).strip()
        options_list = re.split(r'\s*[BCDA]\.\s*', options_text)[1:]
        candidates = []
        for i, option_text in enumerate(options_list, start=65):
            label = chr(i)
            text = option_text.strip()
            candidates.append({
                "label": label,
                "text": text,
                "isRight": 0
            })
        question_info['candidates'] = candidates

    answer_match = re.search(r'答案：([ABCD])', question_block)
    if answer_match and 'candidates' in question_info:
        correct_label = answer_match.group(1)
        for candidate in question_info['candidates']:
            if candidate['label'] == correct_label:
                candidate['isRight'] = 1

    return question_info


class GetArticle:
    def __init__(self):
        self.m = MySQLUploader()
        self.ds = DS()

        self.callback_url_dict = {}
        self.real_ip_dict = {}
        self.demo_name = {}

        self.punctuation = [",", ".", "!", "?", ":", ";", '"', "–", "_", "-", "...", "......"]
        all_exchange_words.update(self.punctuation)

    def parser_insert_to_mysql(self, resp_result):
        for single_article in resp_result['articles']:

            article = single_article['body']
            article_json = json.dumps(single_article)
            difficult_value = find_interval(get_article_difficulty(article))
            if not difficult_value:
                logger.error("文章难度等级为0；")
            sql = "INSERT INTO spring_bamboo_article (article_json,difficult_level) VALUES (%s,%s)"
            self.m.execute_(sql, (article_json, difficult_value))

    def submit_task(self, words_meaning_list: list, take_count: int, student_stage: int, real_ip: str, demo_name: str):
        """
        words_meaning_ids: 词义id 包含词义ID的数组集合，用于生成文章。- 示例：[110, 111, 112, 113, 114]
        take_count: 取文章数量 （int类型，正常是2篇，最大8篇）
        student_stage: 学段（int类型：1.小学;2.初中;3.高中;）
        demo_name: 项目名称
        """
        task_id = randint(10000000, 99999999)

        words_meaning_str = ";".join([i["spell"] + ":" + i["meaning"] for i in words_meaning_list])
        logger.info(f"生成文章id。task_id:{task_id}。词义组:{words_meaning_str}.")

        self.real_ip_dict[task_id] = real_ip
        self.demo_name[task_id] = demo_name

        try:

            resp_result = self.run_task(words_meaning_list, task_id, take_count, student_stage)
            self.parser_insert_to_mysql(resp_result)
            return resp_result
        except Exception as e:
            err_msg = f"GetArticle提交任务失败{type(e).__name__},{e}"
            log_err_e(e, msg="GetArticle提交任务失败;")

            return err_msg

    def get_article(self, words_meaning_list, student_stage, task_id, take_count) -> dict:
        diffculty_control = {
            1: {"grade": "小学", "article_word_count": 60, "desc_difficulty": "最简单最容易没有难度", "paragraph_count": 1,
                "desc2": "文章整体非常简洁，通俗易懂，适合初学者，刚入门，单词全是最常见的，语句通顺即可。",
                "choice_desc": "选择题难度尽可能简单，但是不要让所有选择题让其直接在文中找到答案，允许1-2个选择题很简单，参考中国小学生水平"},
            2: {"grade": "初中", "article_word_count": 200, "desc_difficulty": "简单、常见、难度低", "paragraph_count": 3,
                "desc2": "文章整体难度适中，大约和中国初中生，中国CET-3，雅思4分这样的难度标准。",
                "choice_desc": "选择题难度适中，但是不要所有选择题让其直接在文中找到答案，参考中国初中生水平，中考标准。"},
            3: {"grade": "高中", "article_word_count": 300, "desc_difficulty": "常见、高中难度的", "paragraph_count": 3,
                "desc2": "文章整体难度适中，大约和中国的高中生，中国CET-4，雅思5分这样的难度标准。",
                "choice_desc": "选择题难度偏难，要有迷惑性，不要出现直接在文中找到答案，参考中国高中生水平，高考标准。"}
        }
        take_count_dict = {0: "", 1: "一", 2: "二", 3: "三", 4: "四", 5: "五", 6: "六", 7: "七", 8: "八"}
        different_cou = take_count_dict.get(take_count, "")

        grade = diffculty_control[student_stage]["grade"]
        select_word_count = diffculty_control[student_stage]["article_word_count"]
        select_diffculty = diffculty_control[student_stage]["desc_difficulty"]
        select_paragraph_count = diffculty_control[student_stage]["paragraph_count"]
        desc2 = diffculty_control[student_stage]["desc2"]
        choice_desc = diffculty_control[student_stage]["choice_desc"]

        shuffle(words_meaning_list)
        words_meaning_str = ";".join([i["spell"] + ":" + i["meaning"] for i in words_meaning_list])

        q = f"""不要与前面{different_cou}篇一样，要不同的场景。你是一名在中国的英语教师，下面我会为你提供一些带中文词义的英语单词，请根据这些单词的中文词义，\
生成带英文标题和中文翻译的考场英语文章，注意这个单词有多个词义时，生成的英语文章一定要用提供的中文词义。并挑选一句复杂的句子和其中文翻译，放入difficultSentences。
提供单词：{words_meaning_str}

要求:
1.必须用提供的这个词义的单词，其他单词使用{select_diffculty}的单词。{desc2}
2.文章中使用提供单词，一定要和提供单词的中文词义匹配，尤其是一词多义时，务必使用提供单词的词义。必须要用提供单词的词义。如果用到的词义与提供单词词义不一致，请不要使用这个单词。
3.生成的文章要求{select_word_count}词左右，可以用\\n\\n字符分段，一般{select_paragraph_count}个段落左右。
4.优先保证文章语句通顺，意思不要太生硬。不要为了使用特定的单词，造成文章语义前后不搭，允许不使用个别词义。
5.回复json，格式：{{"title":英文标题,"english":英语文章,"chinese":中文翻译,"difficultSentences": [
    {{
        "english": "",
        "chinese": ""
    }}
]}}
"""
        try:
            real_ip = self.real_ip_dict[task_id]
            demo_name = self.demo_name[task_id]
            r_json = json.loads(
                self.ds.get_article(q, temperature=1, json_resp=True, real_ip=real_ip, demo_name=demo_name, max_tokens=8000))

            r_json["body"] = r_json["title"] + "\n\n" + r_json["english"]
            del r_json["title"]

            q_choice_question = f"""你是一名在中国的{grade}英语教师，下面我会为你提供一篇英语短文，请根据短文设计4个选择题.
短文:{r_json["english"]}

###要求:
1. 生成的选择题不要让学生从短文中直接找到答案，可以混淆，最好让学生推理或排除获得正确答案。用词可以{select_diffculty}，出题要参考中国的中考高考。
2.{choice_desc}
3. 每个选择题之间间隔两行，回复格式如下：

问题: 英语的选择题问题1文本；
解析: 中文的选择题答案解析；
选项：A. 选项1  B. 选项2  C. 选项3  D. 选项4
答案：B

其他几个选择题依次类推
"""

            resp_text = self.ds.get_article(q_choice_question, temperature=1, real_ip=real_ip, demo_name=demo_name, max_tokens=8000)
            questions = resp_text.strip().split('\n\n')

            parsed_questions = [parse_question(q) for q in questions]

            json_data = {"questions": parsed_questions}

            allWordAmount = 0
            allWordAmount += len(split_text_to_word(r_json["english"]))
            for i in json_data["questions"]:
                count_trunk = len(split_text_to_word(i["trunk"]))
                count_candidates = sum([len(split_text_to_word(ii["text"])) for ii in i["candidates"]])
                allWordAmount += count_trunk
                allWordAmount += count_candidates

            return {**r_json, **json_data, "allWordAmount": allWordAmount}
        except json.decoder.JSONDecodeError:
            logger.error("gpt生成文章回复json格式化错误")
        except Exception as e:
            logger.error(f"gpt生成文章回复其他错误.{type(e).__name__} {e}")

    def run_get_article_task(self, words_meaning_list, task_id, take_count, student_stage) -> dict:
        """
        :param words_meaning_list: 数据库内查出来的单词和词义的列表
        :param task_id: 任务id
        :param take_count: 文章数量
        :param student_stage: 学段标识，整型
        :return:
        """
        futures = []
        for i in range(take_count):
            futures.append(pool_executor.submit(self.get_article, words_meaning_list, student_stage, task_id, take_count))
        wait(futures)
        return_json = {"articles": []}
        for t in futures:
            return_json["articles"].append(t.result())
        return return_json

    def run_task(self, words_meaning_list, task_id, take_count, student_stage):
        try:
            outside_json = self.run_get_article_task(words_meaning_list, task_id, take_count, student_stage)
            logger.success(f"文章3-DeepSeek文章任务完成。taskid：{task_id}")
            return outside_json
        except Exception as e:
            logger.error(f"{type(e).__name__} {e}")
        finally:
            self.real_ip_dict.pop(task_id)
            self.demo_name.pop(task_id)