Merge branch 'refs/heads/feature/ISSUE-109' into staging

app/main.py

@@ -4,7 +4,7 @@
 from fastapi import FastAPI
 from starlette.exceptions import HTTPException
 
-from app.router.chatbot_article_detail_router import chatbot_article_router
+from app.router.chatbot_article_router import chatbot_article_router
 from app.config.exception_handler import exception_handler, http_exception_handler
 from app.config.middlewares.request_response_logging_middle_ware import (
     LoggingMiddleware,

app/model/crawled_article.py

@@ -24,12 +24,6 @@ class Articles(Base):
     category = Column(CHAR(255), nullable=True)
     published_at = Column(DateTime, nullable=True)
     image_url = Column(String, nullable=True)
-    probability_issue_finder = Column(Integer, nullable=True)
-    probability_lifestyle_consumer = Column(Integer, nullable=True)
-    probability_entertainer = Column(Integer, nullable=True)
-    probability_tech_specialist = Column(Integer, nullable=True)
-    probability_professionals = Column(Integer, nullable=True)
-
 
 @event.listens_for(Articles, "before_update", propagate=True)
 def update_timestamp(mapper, connection, target):  # pylint: disable=unused-argument

app/model/prompt/simple_article/2024-07-14.txt

@@ -4,18 +4,18 @@
 다음은 json 형식의 예시이다:
 {{
   "title": "MZ세대가 흥미를 끌만한 기사 제목(한국어)",
-  "content": "기사 본문 (한국어). 단, 경제 기사의 독자층이 경제 지식이 부족한 20대 초반인 것을 고려하여 적당한 이모지를 사용하여 친근하고 간결하게 설명할 것. 문단은 \n으로 구분할 것.",
+  "content": "기사 본문 (한국어). 단, 경제 기사의 독자층이 경제 지식이 부족한 20대 초반인 것을 고려하여 적당한 이모지를 사용하여 친근하게 재생성할 것. 문단은 \n으로 구분할 것.",
   "phrase": {{"어려웠던 경제 표현들" :  "어려웠던 경제 표현들을 쉽게 바꾼 문구"}}  (예시: {{"환율" : "다른 나라 돈과 우리나라 돈을 교환하는 비율"}}),
   "comment": "기사를 보고 추론할 수 있는 것 1문장을 친구에게 설명하는 듯한 표현으로",
   "category": "Category 중 하나"
 }}
 
 enum Category:
-    ECONOMY_AND_BUSINESS = "경제 및 기업"
-    POLITICS_AND_SOCIETY = "정치 및 사회"
-    TECHNOLOGY_AND_CULTURE = "기술 및 문화"
-    SPORTS_AND_LEISURE = "스포츠 및 여가"
-    OPINION_AND_ANALYSIS = "오피니언 및 분석"
+    ECONOMY_AND_BUSINESS
+    POLITICS_AND_SOCIETY
+    TECHNOLOGY_AND_CULTURE
+    SPORTS_AND_LEISURE
+    OPINION_AND_ANALYSIS
 
 결과는 json 형식이어야 한다.
 

app/model/subscription.py

@@ -7,11 +7,11 @@
 
 
 class MailTypeCategory(Enum):
-    ECONOMY_AND_BUSINESS = "경제 및 기업"
-    POLITICS_AND_SOCIETY = "정치 및 사회"
-    TECHNOLOGY_AND_CULTURE = "기술 및 문화"
-    SPORTS_AND_LEISURE = "스포츠 및 여가"
-    OPINION_AND_ANALYSIS = "오피니언 및 분석"
+    ECONOMY_AND_BUSINESS = "ECONOMY_AND_BUSINESS"  # 경제와 비즈니스
+    POLITICS_AND_SOCIETY = "POLITICS_AND_SOCIETY"  # 정치와 사회
+    TECHNOLOGY_AND_CULTURE = "TECHNOLOGY_AND_CULTURE"  # 기술과 문화
+    SPORTS_AND_LEISURE = "SPORTS_AND_LEISURE"  # 스포츠와 여가
+    OPINION_AND_ANALYSIS = "OPINION_AND_ANALYSIS"  # 의견과 분석
 
 
 class Subscription(Base):

app/rag_lang_chain/langchain_applied.py

@@ -6,6 +6,7 @@
 from langchain.schema import Document
 from langchain_core.messages import HumanMessage
 from langchain_openai import ChatOpenAI
+from pydantic.v1 import BaseModel
 
 from app.config.loguru_config import logger
 from app.rag_lang_chain.chromadb_manager import ChromaDBManager
@@ -15,9 +16,9 @@
 )
 
 
-class RagAppliedResult:
+class RagAppliedResult(BaseModel):
     result_text: str
-    related_documents: List[Union[Document, dict]]
+    related_documents: List[Document]
 
 
 async def request_rag_applied_openai(
@@ -34,11 +35,7 @@ async def request_rag_applied_openai(
     )
 
     # Step 1: Google Custom Search API를 사용하여 관련 정보 수집
-    google_results = await google_cse_retriever.retrieve(
-        original_text
-    )  # FIXME: 왜 GoogleCSERetriever를 사용하는가? # pylint: disable=fixme
-    if not google_results:
-        raise HTTPException(status_code=404, detail="No results found from Google.")
+    google_results = await google_cse_retriever.retrieve(original_text)
 
     # Step 2: 검색 결과를 벡터화하고 ChromaDB에 저장
     chroma_db_manager = ChromaDBManager()
@@ -49,14 +46,17 @@ async def request_rag_applied_openai(
     additional_info = await search.aget_relevant_documents(original_text, num_results=3)
 
     # Step 4: 프롬프트 생성(원문 + 검색 결과 + 추가 정보)
-    rag_applied_prompt = await create_rag_applied_prompt(
-        original_prompt=system_prompt, relevant_info=search_results + additional_info
-    )
+    rag_applied_prompt = system_prompt
+    if search_results:
+        rag_applied_prompt = await create_rag_applied_prompt(
+            original_prompt=system_prompt,
+            relevant_info=search_results + additional_info,
+        )
 
     # Step 5: OpenAI 요청 결과 반환
     try:
         search_llm = ChatOpenAI(
-            temperature=0, model="gpt-4", max_tokens=1500, api_key=openai_api_key
+            temperature=0, model="gpt-4o", max_tokens=1500, api_key=openai_api_key
         )
         response = await search_llm.agenerate(
             messages=[[HumanMessage(rag_applied_prompt)]]
@@ -70,11 +70,9 @@ async def request_rag_applied_openai(
 
     logger.info(f"Response: {response.generations[0][0].text}")
 
-    # response.generations[0][0].text
-
     return RagAppliedResult(
         result_text=response.generations[0][0].text,
-        related_documents=search_results + additional_info,
+        related_documents=search_results,
     )
 
 

app/repository/crawled_article_crud.py

@@ -21,22 +21,6 @@ async def get(self, pk: int, session: AsyncSession):
             )
         return article
 
-    async def set_interest_type(
-            self, pk:int, interest_types : List[int], session: AsyncSession
-    ):
-        repository = get_repository(Articles)(session)
-        return await repository.update_by_pk(
-            pk = pk,
-            data = {
-                'probability_issue_finder': interest_types[0],
-                'probability_lifestyle_consumer': interest_types[1],
-                'probability_entertainer': interest_types[2],
-                'probability_tech_specialist': interest_types[3],
-                'probability_professionals': interest_types[4]
-            }
-
-        )
-
     async def get_all(self, session: AsyncSession):
         repository = get_repository(Articles)(session)
         return await repository.filter()

app/router/chatbot_article_detail_router.py → app/router/chatbot_article_router.py

@@ -1,26 +1,30 @@
-from fastapi import APIRouter
+from fastapi import APIRouter, Depends
 from pydantic import BaseModel
 
-from app.service.chatbot_article_detail_service import request_rag_applied_openai
+from sqlalchemy.ext.asyncio import AsyncSession
+from app.database.session import get_db_session
+from app.service.chatbot_article_service import request_rag_applied_openai
 from app.utils.generic_response import GenericResponseDTO
 
 chatbot_article_router = APIRouter()
 
 # 사용자 요청
 class GenerateDetailArticleRequestDTO(BaseModel):
-    news_content: str
+    id : int
     prompt: str
 
 
 @chatbot_article_router.post(
-    "/chatbot-article-detail", response_model=GenericResponseDTO
+    "/chatbot/article", response_model=GenericResponseDTO
 )
 async def chatbot_article_detail_(
         request: GenerateDetailArticleRequestDTO,
+        session: AsyncSession = Depends(get_db_session)
 ):
     rag_applied_result = await request_rag_applied_openai(
-        original_text=request.news_content,
-        system_prompt=request.prompt
+        news_id =request.id,
+        system_prompt=request.prompt,
+        session = session
     )
 
     return GenericResponseDTO(

app/service/article_manage_service.py

@@ -1,11 +1,14 @@
 from datetime import datetime
 from typing import List
+
 from sqlalchemy.ext.asyncio import AsyncSession
+
 from app.model.article_publisher import Publisher
 from app.model.crawled_article import Articles
 from app.model.subscription import MailTypeCategory
 from app.repository.crawled_article_crud import CrawledArticleRepository
 
+
 class ArticleManageService:
     async def create_article(
         self,
@@ -18,7 +21,7 @@ async def create_article(
         phrase: dict,
         comment: str,
         image_url: str,
-        published_at: str,
+        published_at: datetime,
         category: MailTypeCategory,
         session: AsyncSession,
     ) -> Articles:
@@ -31,11 +34,10 @@ async def create_article(
                 simple_title=simple_title,
                 simple_content=simple_content,
                 comment=comment,
-                published_at=datetime.strptime(published_at, '%Y-%m-%dT%H:%M:%S'),
+                published_at=published_at,
                 image_url=image_url,
                 category=category.name,
                 phrase=phrase,
-                probability_issue_finder=-1
             ),
             session=session,
         )

app/service/chatbot_article_detail_service.py → app/service/chatbot_article_service.py

@@ -3,14 +3,17 @@
 import aiohttp
 from fastapi import HTTPException
 from langchain_core.messages import HumanMessage
+from langchain_core.outputs import LLMResult
 from langchain_openai import ChatOpenAI
+from sqlalchemy.ext.asyncio import AsyncSession
 
 from app.config.loguru_config import logger
 from app.rag_lang_chain.chromadb_manager import ChromaDBManager
 from app.rag_lang_chain.google_cse_retriver import (
     AsyncGoogleSearchAPIWrapper,
     GoogleCSERetriever,
 )
+from app.service.article_manage_service import ArticleManageService
 
 from langchain.schema import Document
 
@@ -40,8 +43,9 @@ def to_dict(self) -> Dict:
 
 
 async def request_rag_applied_openai(
-        original_text: str,  # OriginalText: 기사 원문(Google Custom Search에 보낼 용도)
-        system_prompt: str,  # SystemPrompt: 시스템 프롬프트
+        news_id: int,
+        system_prompt: str,
+        session : AsyncSession
 ) -> Dict:
     openai_api_key = os.getenv("OPENAI_API_KEY")
     google_api_key = os.getenv("GOOGLE_API_KEY")
@@ -52,14 +56,26 @@ async def request_rag_applied_openai(
         api_key=google_api_key, cse_id=google_cse_id
     )
 
+    # Step 0 : 기사 id값에 따른 기사 원문 가져오기
+    article_service = ArticleManageService()
+    article_by_id = await article_service.get_article_by_id(news_id, session)
+    original_text = article_by_id.content
+    if not original_text:
+        raise HTTPException(status_code=404, detail="Article not found.")
+
+
     # Step 1: Google Custom Search API를 사용하여 사용자가 입력한 original_text 관련 정보 전부 수집
     # original_text와 관련된 웹 페이지의 목록을 반환함. 각 웹 페이지는 title(검색 결과 제목), link(웹 페이지 url), snippet(검색 결과의 요약)으로 구성됨.
     google_results = await google_cse_retriever.retrieve(
         original_text
     )
-    if not google_results:
-        raise HTTPException(status_code=404, detail="No results found from Google.")
     logger.info(f"1. Google results: {google_results}")
+    if not original_text:
+        response = await openai_response(openai_api_key, system_prompt)
+        return RagAppliedResult(
+            result_text=response.generations[0][0].text,
+            related_documents=[],
+        ).to_dict()
 
     # Step 2: 검색 결과를 벡터화하고 ChromaDB에 저장
     chroma_db_manager = ChromaDBManager()
@@ -71,38 +87,51 @@ async def request_rag_applied_openai(
     additional_info = await search.aget_relevant_documents(original_text, num_results=3)
     logger.info(f"3. Additional info: {additional_info}")
 
-    # Step 4: 프롬프트 생성(원문 + 검색 결과 + 추가 정보)
-
+    # Step 4: 프롬프트 생성 (원문 + 검색 결과 + 추가 정보)
     rag_applied_prompt = await create_rag_applied_prompt(
-        original_prompt=system_prompt, relevant_info=search_results + additional_info
+        original_prompt=system_prompt, relevant_info=search_results + additional_info, original_text=original_text
     )
 
     # Step 5: OpenAI 요청 결과 반환
+    response = await openai_response(openai_api_key, rag_applied_prompt)
+
+    logger.info(f"최종 Response: {response}")
+
+    return RagAppliedResult(
+        result_text=response.generations[0][0].text,
+        related_documents=search_results + additional_info,
+    ).to_dict()
+
+
+# OpenAI 요청 결과 반환
+async def openai_response(
+        openai_api_key: str,
+        prompt: str
+) -> LLMResult:
     try:
         search_llm = ChatOpenAI(
             temperature=0, model="gpt-4", max_tokens=1500, api_key=openai_api_key
         )
         response = await search_llm.agenerate(
-            messages=[[HumanMessage(rag_applied_prompt)]]
+            messages=[[HumanMessage(prompt)]]
         )
+        return response
     except aiohttp.ClientResponseError as e:
         if e.status == 429:
             raise HTTPException(
                 429, "Too many requests. Please try again later."
             ) from e
         raise HTTPException(500, "Internal Server Error") from e
 
-    logger.info(f"최종 Response: {response}")
-
-    return RagAppliedResult(
-        result_text=response.generations[0][0].text,
-        related_documents=search_results + additional_info,
-    ).to_dict()
 
 
 async def create_rag_applied_prompt(
-    original_prompt: str, relevant_info: List[Union[Document, dict]]
+    original_prompt: str, relevant_info: List[Union[Document, dict]], original_text: str
 ) -> str:
+    # 원문 기사 추가
+    original_prompt += f"\n원문 기사:\n{original_text}\n\n"
+
+    # 관련 정보 추가
     for idx, info in enumerate(relevant_info):
         if isinstance(info, Document):
             title = info.metadata.get("title", "제목 없음")
@@ -113,7 +142,7 @@ async def create_rag_applied_prompt(
             link = info.get("link", "URL 없음")
             snippet = info.get("snippet", "내용 없음")
         original_prompt += (
-            f"\n{idx + 1}. 제목: {title}\n   URL: {link}\n   내용: {snippet}\n"
+            f"관련된 기사 정보들 \n{idx + 1}. 제목: {title}\n   URL: {link}\n   내용: {snippet}\n"
         )
 
     logger.info(f"RAG Applied Prompt: {original_prompt}")

app/service/news_scheduling_service.py

@@ -8,6 +8,7 @@
 from app.config.loguru_config import logger
 from app.database.session import db_session
 from app.model.article_publisher import Publisher
+from app.recommend.recommend_service import RecommendService
 from app.service.article_manage_service import ArticleManageService
 from app.service.simple_article_service import process_generate_article_by_url
 
@@ -87,7 +88,6 @@ async def run_crawl_and_store(session: AsyncSession):
         )
 
 
-
 async def schedule_task():
     while True:
         now = datetime.now()