systex_phase2/sql_qa_test.py

import csv
import time
from langchain.callbacks import get_openai_callback
from ai_agent import main
from systex_app import save_history


def agent(question):
    print(question)
    start = time.time()
    with get_openai_callback() as cb:
        result = main(question)
        answer = result["generation"]
    processing_time = time.time() - start
    save_history(question, answer, cb, processing_time)
    
    if "test@systex.com" in answer:
        answer = "很抱歉，目前我無法回答您的問題，請將您的詢問發送至 test@systex.com 以便獲得更進一步的幫助，謝謝。"
    print(answer)
    return result

places = {
    '高雄': ["高雄廠", "高雄總部及運通廠", "運通廠", "總部", "高雄總部", "高雄"], 
    '台北': ["台北廠", "台北辦事處", "台北", "臺北"],
    '廣興': ["廣興廠", "昆山廠", "廣興"],
    '北海': ["北海建準廠", "北海廠", "立準廠", "立準", "北海"],
    '菲律賓': ["菲律賓建準廠", "菲律賓"],
    '印度': ["India", "印度"],
    '美國': ["Inc", "美國"],
    '法國': ["SAS", "法國"],
}
items = ['直接排放', '能源間接排放', '運輸間接排放', '組織使用產品間接排放', '使用來自組織產品間接排放',
        '類別1', '類別2', '類別3', '類別4', '類別5', '類別6',
        '固定燃燒', '移動燃燒', '製程排放', '逸散排放', '土地利用', 
        '外購電力', '外購能源', 
        '上游運輸', '下游運輸', '員工通勤', '商務旅行', '訪客運輸', 
        '購買產品', '外購燃料及能資源', '資本貨物', '上游租賃', '廢棄物處理', '廢棄物清運', '其他委外業務', 
        '產品加工', '產品使用', '產品最終處理', '下游租賃', '投資排放', 
        '其他間接排放']
items2 = ['綠電', "灰電", "自產電力", "外購電力"]
times = ["2022", "2023", "去年", "前年"]
import random
import math

question_list = []
for place in places.keys():
    print(place)
    loc_random = random.sample(places[place], k=math.ceil(len(places[place])/2))
    print(loc_random)
    for loc in loc_random:
        # item_random = random.sample(items, k=10)
        item_random = items2
        for item in item_random:
            year_time = random.choice(times)
            # questions_format = [f"建準{place}{year_time}的{item}是多少", f"{place}{year_time}的{item}排放量", f"{year_time}{place}的{item}是多少", 
            #                     f"請問{year_time}的{place}{item}排放量是多少？", f"{year_time}建準{place}{item}"]
            questions_format = [f"建準{place}{year_time}的{item}是多少", f"{place}{year_time}的{item}使用量", f"{year_time}{place}的{item}是多少", 
                                f"請問{year_time}的{place}{item}使用量是多少？", f"{year_time}建準{place}{item}"]
            
            questions = random.sample(questions_format, k=2)
            # question_list.extend(questions)
            for question in questions:
                print(question)
                result = agent(question)
                
                labels = result.keys()
                with open("results2.csv", "a") as f:
                    writer = csv.DictWriter(f, fieldnames=labels)
                    writer.writeheader()
                    writer.writerow(result)