promptSim/text_analysis/views.py_20240607

# coding:utf8
import os, sys
import io
from jsonpath_ng import jsonpath, parse
import uuid
sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf8')
cur_dir = os.path.dirname(os.path.abspath(__file__)) or os.getcwd()
par_dir = os.path.abspath(os.path.join(cur_dir, os.path.pardir))
sys.path.append(cur_dir)
sys.path.append(par_dir)
import json
from django.http import HttpResponse
from text_analysis.tools import to_kafka
from django.views.decorators.csrf import csrf_exempt
from log_util.set_logger import set_logger
from datetime import datetime, timedelta
logging = set_logger('logs/results.log')
import traceback
import queue
import requests
from langchain.embeddings.sentence_transformer import SentenceTransformerEmbeddings
from text_analysis.tools.tool import parse_data,promptPro
from text_analysis.chroma1 import LangChainChroma
import time
from datetime import datetime
import os
from langchain.vectorstores import Chroma

# 任务队列
global task_queue
task_queue = queue.Queue()
# LC = LangChainChroma()


@csrf_exempt
def promptSim(request):
    if request.method == 'POST':
        try:
            raw_data = json.loads(request.body)
            task_queue.put(raw_data)
            return HttpResponse(json.dumps({"code": 1, "msg": "请求正常！"}, ensure_ascii=False))
        except:
            logging.error(traceback.format_exc())
            return HttpResponse(json.dumps({"code": 0, "msg": "请求json格式不正确！"}, ensure_ascii=False))
    else:
        return HttpResponse(json.dumps({"code": 0, "msg": "请求方式错误，改为post请求"}, ensure_ascii=False))


def upload():
    while True:
        try:
            if task_queue.qsize() > 0:
                logging.info("取任务队列长度{}".format(task_queue.qsize()))
                raw_data = task_queue.get()
                output=raw_data["output"]
                res_tmp={key: "" for key in output}
                if "id" in res_tmp.keys():
                    res_tmp["id"]=str(uuid.uuid4())
                logging.info("任务数据为：{}".format(raw_data))
                topn=parse_data(raw_data,raw_data["input"]["topn"])
                prompt=parse_data(raw_data,raw_data["input"]["prompt"])
                fieldName=parse_data(raw_data,raw_data["input"]["fieldName"])
                vector_db=LangChainChroma(fieldName)
                docs=vector_db.search(prompt,topn)
                vector_db.db_close()
                logging.info("向量数据库搜索的相似上下文：{}".format(docs))
                #组装，最长字符5W
                res=promptPro(prompt,docs)
                logging.info("生成的上下文：{}".format(res))
                res_tmp["promptRes"]=res
                res_tmp_json = json.dumps(res_tmp, ensure_ascii=False)
                raw_data["result"] = {"successCode": "1", "errorLog": "", "results": res_tmp_json}
                raw_data["result"]["status"] = 1
                raw_data["result"]["message"] = "成功"
                logging.info("结果数据为：{}".format(raw_data))
                to_kafka.send_kafka(raw_data, logging)
            else:
                # 暂无任务，进入休眠
                time.sleep(10)
        except:
            raw_data["result"]={}
            raw_data["result"]["successCode"] = "0"
            raw_data["result"]["errorLog"] = traceback.format_exc()
            raw_data["result"]["status"] = 2
            raw_data["result"]["message"] = "异常"
            raw_data["result"]["results"] = json.dumps(res_tmp, ensure_ascii=False)
            logging.error(traceback.format_exc())
            to_kafka.send_kafka(raw_data, logging)