ChatGPT로 은행에서는 어떤 애플리케이션을 만들어 볼 수 있을까?

import openai
import re

from src.db_utils import *
from src.prompt import *
from config.settings import *

openai.api_key = OPENAI_KEY
class MKTPhraseGenerator:
def __init__(self):

self.prompt = PROMPT_DICT[“MKT_PHRASE_PRMPT”]

self.basic_prod_dic = {
‘aptloan’: “아파트 담보대출”,
‘codek_freeaccount’: “자유적금”,
‘coin_info’: “가상자산(비트코인) 시세정보”,
‘hiteenpromotion’: “중/고등학생 체크카드”,
‘homelearn_event’: “초등학생 학습지”,
‘togetheraccount’: “모임통장”
}

self.error_template = {
‘aptloan’: {‘text1’: ‘큰 돈 새는 주담대 이자 이제는 이자도 절약해요 부자 솔루션’, ‘text2’: ‘- 절약 비교해보면 케뱅 아파트담보대출’},
‘codek_freeaccount’: {‘text1’: ‘단순하고 쉬운게 좋다면? ‘, ‘text2’: ‘코드K 자유적금이 딱이에요 복잡한 우대조건 필요 없어요’},
‘coin_info’: {‘text1’: ‘비트코인 시세 조회 어디서 하세요?’, ‘text2’: ‘여기서도 간편하게 조회가 가능해요!’},
‘hiteenpromotion’: {‘text1’: ‘우리 아이 버스•지하철 교통비 500원씩 싸게 타는 방법이 있어요’, ‘text2’: ‘Hi teen 자세히 알아보기’},
‘homelearn_event’: {‘text1’: ‘우리 아이 성적에 고민이 많으신 부모님들!’, ‘text2’: ‘전교 1등 노하우 알려드릴게요!’},
‘togetheraccount’: {‘text1’: ‘같이 모으는 돈에도’, ‘text2’: ‘최고 연 10%(세전) 이자 받기’}
}
pass

def generate_mkt_phrase(self, prod_nm, name, prompt):

“””
Input:
prod_nm: str
name: str
prompt: str
Return:
text1: str ex) 큰 돈 새는 주담대 이자 이제는 이자도 절약해요 부자 솔루션
text2: str ex) – 절약 비교해보면 케뱅 아파트담보대출
“””

response = self._generate_chat_completion(prompt)

print(response)

try:
text1, text2 = self.preprocess_response(response, name)
except:
# error handle
error_handle_texts = self.error_template[prod_nm]
text1, text2 = error_handle_texts[‘text1’], error_handle_texts[‘text2’]

return text1, text2

def preprocess_response(self, response, name):
“””
Input:
response: str
Return:
text1: str
text2: str
“””

def _preprocess_pipeline(response):

pipe_cnt = response.split(‘|’)

if len(pipe_cnt) == 2:
text1, text2 = response.split(“|”)
return text1, text2

else:
texts = response.split(‘|’)
text1 = texts[0]
text2 = ” “.join(texts[1:])
return text1, text2

def _preprocess_text(text, name):
# regex
text = re.sub(‘[0-9]위.’, ”, text)
text = re.sub(‘[0-9]위’, ”, text)

# replace
text = text.replace(‘결과:’, ”)
text = text.replace(‘”‘, ”)
text = text.replace(‘\n’, ”)
text = text.replace(‘{CUST_FULL_NM}’, name)
text = text.replace(‘(CUST_FULL_NM)’, name)
text = text.replace(‘[CUST_FULL_NM]’, name)
text = text.replace(‘{고객}’, name)
text = text.replace(‘(고객)’, name)
text = text.replace(‘[고객]’, name)
text = text.replace(‘고객’, name)
text = text.replace(‘(첫 번째 문장)’, “”)
text = text.replace(‘(두 번째 문장)’, “”)

return text

text1, text2 = _preprocess_pipeline(response)
text1, text2 = _preprocess_text(text1, name), _preprocess_text(text2, name)
text1, text2 = text1.strip(), text2.strip()

return text1, text2

def _generate_chat_completion(self, prompt):

response = openai.ChatCompletion.create(
model=”gpt-3.5-turbo”,
messages=[
{
“role”: “user”,
“content”: prompt
}
]
)

response = response[“choices”][0][“message”][“content”]

return response

def _get_basic_prod(self, prod_nm):
return self.basic_prod_dic[prod_nm]

def generate_prompt(self, prod_nm, geng, ageg, jobseg):

basic_prod_nm = self._get_basic_prod(prod_nm)
prod_infos, _ = self._search_prod_infos(prod_nm)

prod_info = prod_infos[‘prod_info’]
prod_promotion_info = prod_infos[‘prod_promotion_info’]
prod_ir = prod_infos[‘prod_ir’]

best_mkt_phrases, _ = self._search_best_mkt_phrases(prod_nm, geng, ageg, jobseg)

prompt = self.prompt.format(
basic_prod_nm=basic_prod_nm,
best_mkt_phrases=best_mkt_phrases,
prod_info=prod_info,
prod_promotion_info=prod_promotion_info,
prod_ir=prod_ir,
ageg=ageg,
geng=geng,
jobseg=jobseg
)

return prompt

def _search_prod_infos(self, prod_nm):
“””
Input:
prod_nm: str ex) ‘aptloan’
Return:
best_mkt_phrases: dict
“””
db = get_db_connection(host=SQL_HOST, user=SQL_USERNAME, passwd=SQL_PASSWD, db=SQL_SCHEMA_NAME)
try:
with db.cursor() as curs:
curs.execute(‘SET SESSION TRANSACTION ISOLATION LEVEL READ UNCOMMITTED’)
db.commit()

sql = f”””
SELECT prod_nm, prod_info, prod_promotion_info, prod_ir
FROM prd_info
WHERE
prod_nm = ‘{prod_nm}’
“””

print(sql)

curs.execute(sql)
rows = curs.fetchall()
row = rows[0]

prod_nm = row[0]
prod_info = row[1].replace(‘\\n’,’\n’)
prod_promotion_info = row[2].replace(‘\\n’,’\n’)
prod_ir = row[3].replace(‘\\n’,’\n’)

prod_infos = {
‘prod_nm’ : prod_nm,
‘prod_info’: prod_info,
‘prod_promotion_info’: prod_promotion_info,
‘prod_ir’: prod_ir
}

return prod_infos, True

except:

prod_infos = {
‘prod_nm’ : ”,
‘prod_info’: ”,
‘prod_promotion_info’: ”,
‘prod_ir’: ”
}
return prod_infos, False

def _search_best_mkt_phrases(self, prod_nm, geng, ageg, jobseg):
“””
Input:
prod_nm: str ex) ‘aptloan’
gegng: str ex) ‘여성’
ageg: str ex) ’20대’
jpbseg str ex) ‘전문직’
Return:
best_mkt_phrases: str
“””
db = get_db_connection(host=SQL_HOST, user=SQL_USERNAME, passwd=SQL_PASSWD, db=SQL_SCHEMA_NAME)
try:
with db.cursor() as curs:
curs.execute(‘SET SESSION TRANSACTION ISOLATION LEVEL READ UNCOMMITTED’)
db.commit()

sql = f”””
SELECT mkt_phrase
FROM mkt_phrase
WHERE
prod_nm = ‘{prod_nm}’
and geng = ‘{geng}’
and ageg = ‘{ageg}’
and jobseg = ‘{jobseg}’
ORDER BY ctr desc
LIMIT 5;
“””

print(sql)

curs.execute(sql)
rows = curs.fetchall()

best_mkt_phrases = “”
for idx, row in enumerate(rows):
mkt_phrase = row[0]
best_mkt_phrases += f”{idx + 1}위. {mkt_phrase}\n”

return best_mkt_phrases, True

except:
best_mkt_phrases = “”
return best_mkt_phrases, False

if __name__ == ‘__main__’:

# user info
user_info = {
‘prod_nm’: ‘aptloan’,
‘name’: ‘kbank’,
‘geng’: ‘여성’,
‘ageg’: ’20대’,
‘jobseg’: ‘전문직’
}

prod_nm = user_info[‘prod_nm’]
name, geng, ageg, jobseg = user_info[‘name’], user_info[‘geng’], user_info[‘ageg’], user_info[‘jobseg’]

# define mkt_phrase_generator
mkt_phrase_generator = MKTPhraseGenerator()

# generate prompt
prompt = mkt_phrase_generator.generate_prompt(prod_nm, geng, ageg, jobseg)

# get openai response
text1, text2 = mkt_phrase_generator.generate_mkt_phrase(prod_nm, name, prompt)

# output
print(text1, text2)