Merge pull request #52 from rimgosu/main

feat: tts, gpt

.gitignore

@@ -1,4 +1,6 @@
 /fastapi-env/
+/temp/
+/temp/voice_temp
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]

app/api/api_v1/endpoints/log.py

@@ -3,14 +3,14 @@
 from typing import List
 
 from app.database import get_db
-from app.services.log_service import (create_log_service, get_log_service, get_logs_by_gree_service, get_logs_service, delete_log_service)
-from app.schemas.LogDto import CreateLogDto, LogResponseDto
+from app.services.log_service import (create_usertalk_log_service, get_log_service, get_logs_by_gree_service, get_logs_service, delete_log_service)
+from app.schemas.LogDto import CreateUserTalkLogDto, LogResponseDto
 
 router = APIRouter()
 
 @router.post("/", response_model=LogResponseDto)
-async def create_log(log_dto: CreateLogDto, db: AsyncSession = Depends(get_db)):
-    return await create_log_service(db, log_dto)
+async def create_log(log_dto: CreateUserTalkLogDto, db: AsyncSession = Depends(get_db)):
+    return await create_usertalk_log_service(db, log_dto)
 
 @router.get("/{log_id}", response_model=LogResponseDto)
 async def read_log(log_id: int, db: AsyncSession = Depends(get_db)):

app/crud/crud_log.py

@@ -4,11 +4,12 @@
 from app.models.models import Log  # 모델 파일 경로에 따라 수정해야 할 수 있음
 
 # 로그 생성
-async def create_log(db: AsyncSession, gree_id: int, log_type, content: str):
+async def create_log(db: AsyncSession, gree_id: int, log_type, content: str, voice_url: str):
     db_log = Log(
         gree_id=gree_id,
         log_type=log_type,
         content=content,
+        voice_url=voice_url,
         register_at=datetime.now()
     )
     db.add(db_log)

app/models/enums.py

@@ -38,3 +38,12 @@ class FileTypeEnum(PyEnum):
 class EmotionTypeEnum(PyEnum):
     HAPPY = "HAPPY"
     UNHAPPY = "UNHAPPY"
+
+@unique
+class VoiceTypeEnum(PyEnum):
+    ALLOY = 'alloy'
+    ECHO = 'echo'
+    FABLE = 'fable'
+    ONYX = 'onyx'
+    NOVA = 'nova'
+    SHIMMER = 'shimmer'

app/models/models.py

@@ -2,7 +2,7 @@
 from sqlalchemy import Column, Integer, String, DateTime, Enum, ForeignKey, Boolean
 from sqlalchemy.orm import relationship
 from sqlalchemy.ext.declarative import declarative_base
-from app.models.enums import RoleEnum, StatusEnum, GradeEnum, LogTypeEnum, FileTypeEnum
+from app.models.enums import RoleEnum, StatusEnum, GradeEnum, LogTypeEnum, FileTypeEnum, VoiceTypeEnum
 from datetime import datetime
 
 Base = declarative_base()
@@ -34,6 +34,8 @@ class Gree(Base):
     prompt_mbti = Column(String(255))
     status = Column(Enum(StatusEnum))
     isFavorite = Column(Boolean, default=False)
+    # 그리는 프롬프트 엔지니어링을 통해 다음과 같이 TTS의 목소리가 결정되어야한다.
+    voice_type = Column(Enum(VoiceTypeEnum), default=VoiceTypeEnum.ALLOY)
     register_at = Column(DateTime, nullable=False, default=datetime.now())
 
     member = relationship("Member", back_populates="gree")
@@ -62,6 +64,8 @@ class Log(Base):
     gree_id = Column(Integer, ForeignKey('gree.gree_id'), nullable=False)
     log_type = Column(Enum(LogTypeEnum), nullable=False)
     content = Column(String(1000))
+    # "GREE_TALK"일 경우 음성데이터가 Azure에 업로드 되어야한다.
+    voice_url = Column(String(255))
     register_at = Column(DateTime, nullable=False, default=datetime.now())
 
     gree = relationship("Gree", back_populates="log")

app/schemas/LogDto.py

@@ -4,11 +4,17 @@
 from app.models.enums import LogTypeEnum
 
 # 생성을 위한 DTO
-class CreateLogDto(BaseModel):
+class CreateUserTalkLogDto(BaseModel):
     gree_id: int
     log_type: LogTypeEnum
     content: str
 
+class CreateGreeTalkLogDto(BaseModel):
+    gree_id: int
+    log_type: LogTypeEnum
+    content: str
+    voice_url: str # azure
+
 # 응답을 위한 DTO
 class LogResponseDto(BaseModel):
     id: int

app/schemas/gree.py

@@ -21,6 +21,6 @@ class Gree(BaseModel):
     prompt_mbti: Optional[str] = None
     status: StatusEnum
     isFavorite: bool
-
+    
     class Config:
         from_attributes = True

app/services/ai_service.py

@@ -1,11 +1,16 @@
+import os
+import uuid
+import aiohttp
 import openai
 from fastapi import HTTPException
 from sqlalchemy.ext.asyncio import AsyncSession
+from azure.storage.blob import BlobServiceClient, ContentSettings
 
 from app.crud.crud_gree import crud_get_gree_by_id_only
+from app.models.enums import VoiceTypeEnum
 from app.schemas.ChatDto import ChatRequestDto
-from app.schemas.LogDto import CreateLogDto
-from app.services.log_service import create_log_service
+from app.schemas.LogDto import CreateGreeTalkLogDto, CreateUserTalkLogDto
+from app.services.log_service import create_greetalk_log_service, create_usertalk_log_service
 
 # chat 테스트를 위한 서비스이다.
 async def chat_with_openai_test_service(chat_request):
@@ -47,10 +52,13 @@ async def chat_with_openai_test_service(chat_request):
 # 1. gree_id로 gree를 하나 조회한다. ✅
 # 2. gree의 정보를 토대로 chat_request에 있는 정보를 채워넣는다. ✅
 # 3. 즉 엔드포인트에서 받아야 할 정보는 gree_id, message 이 둘이면 된다. ✅
-# 4. gree_id, message로 입력받은 것은 "USER_TALK" Log로 남아야 한다.
-# 5. gree_id, message로 출력되는 것은 "GREE_TALK" Log로 데이터베이스에 저장되어야 한다.
+# 4. gree_id, message로 입력받은 것은 "USER_TALK" Log로 남아야 한다. ✅
+# 5. gree_id, message로 출력되는 것은 "GREE_TALK" Log로 데이터베이스에 저장되어야 한다. ✅
+# 6. "GREE_TALK" 메시지는 TTS과정을 거쳐야한다.
+# 7. TTS를 거친 메시지는 Azure에 저장되어야한다.
+# 8. Azure에 저장된 URL은 Log에 저장되어야한다.
+# 9. GREE의 성격을 지정할때(Update Gree) Gree객체의 VoiceTypeEnum이 지정되어야한다.
 async def chat_with_openai_service(db: AsyncSession, chat_request: ChatRequestDto):
-
     gree = await crud_get_gree_by_id_only(db, chat_request.gree_id)
     system_message = (
         "이 대화는 한국어로 진행됩니다. 모든 응답은 한국어로 제공되어야 합니다. "
@@ -67,13 +75,13 @@ async def chat_with_openai_service(db: AsyncSession, chat_request: ChatRequestDt
         f"당신은 {gree.prompt_gender}, {gree.prompt_age}살, 이름은 {gree.gree_name}, MBTI는 각각의 성향이 강하게 나타나는 {gree.prompt_mbti}입니다."
     )
 
-    createUserLogDto = CreateLogDto(
+    createUserLogDto = CreateUserTalkLogDto(
         gree_id=gree.id,
         log_type='USER_TALK',
         content=chat_request.message
     )
 
-    user_talk = await create_log_service(db ,createUserLogDto)
+    user_talk = await create_usertalk_log_service(db ,createUserLogDto)
 
     try:
         completion = openai.ChatCompletion.create(
@@ -87,16 +95,82 @@ async def chat_with_openai_service(db: AsyncSession, chat_request: ChatRequestDt
                 {"role": "user", "content": chat_request.message}
             ]
         )
+        gree_talk= completion.choices[0].message.content
+
+        mp3_path = await text_to_speech(gree.voice_type, gree_talk)
+        voice_url_azure= await upload_mp3_azure(mp3_path)
 
-        createGptLogDto = CreateLogDto(
+        createGptLogDto = CreateGreeTalkLogDto(
             gree_id=gree.id,
             log_type='GREE_TALK',
-            content=completion.choices[0].message.content
+            content=gree_talk,
+            voice_url=voice_url_azure
         )
 
-        gpt_talk = await create_log_service(db ,createGptLogDto)
+        gpt_talk = await create_greetalk_log_service(db ,createGptLogDto)
 
         return {"user_talk": user_talk, "gpt_talk": gpt_talk}
     except Exception as exc:
         print(f"An error occurred: {exc}")
         raise HTTPException(status_code=500, detail="An error occurred while processing your request.")
+
+
+async def text_to_speech(voice_type: VoiceTypeEnum, gree_talk: str) -> str:
+    url = "https://api.openai.com/v1/audio/speech"
+    OPENAI_API_KEY = os.getenv('OPENAI_API_KEY') # 환경변수에서 API 키를 가져옵니다.
+
+    payload = {
+        "model": "tts-1",
+        "input": gree_talk,
+        "voice": voice_type.value
+    }
+    headers = {
+        "Authorization": f"Bearer {OPENAI_API_KEY}",
+        "Content-Type": "application/json"
+    }
+
+    async with aiohttp.ClientSession() as session:
+        async with session.post(url, json=payload, headers=headers) as response:
+            if response.status == 200:
+                content = await response.read()
+
+                # 지정된 경로에 디렉토리가 없으면 생성
+                dir_path = '/temp/voice_temp/'
+                if not os.path.exists(dir_path):
+                    os.makedirs(dir_path)
+
+                # 파일 저장 경로 변경
+                mp3_path = os.path.join(dir_path, f'{voice_type.value}_speech.mp3')
+
+                with open(mp3_path, 'wb') as f:
+                    f.write(content)
+                return mp3_path
+            else:
+                return "Error: 응답 상태가 200이 아닙니다."
+
+
+async def upload_mp3_azure(mp3_path: str) -> str:
+    try:
+        container_name = "greefile"
+        AZURE_ACCOUNT_KEY = os.getenv("AZURE_ACCOUNT_KEY")
+        if not AZURE_ACCOUNT_KEY:
+            raise HTTPException(status_code=500, detail="Azure account key is not set in environment variables.")
+
+        connection_string = f"DefaultEndpointsProtocol=https;AccountName=greedotstorage;AccountKey={AZURE_ACCOUNT_KEY};EndpointSuffix=core.windows.net"
+        blob_service_client = BlobServiceClient.from_connection_string(connection_string)
+
+        # 파일 이름 추출 및 고유한 파일 이름 생성
+        file_name = os.path.basename(mp3_path)
+        unique_file_name = f"logVoices/{uuid.uuid4()}_{file_name}"
+
+        blob_client = blob_service_client.get_blob_client(container=container_name, blob=unique_file_name)
+
+        # 파일 업로드
+        with open(mp3_path, "rb") as data:
+            blob_client.upload_blob(data, overwrite=True, content_settings=ContentSettings(content_type='audio/mpeg'))
+
+        # 업로드된 파일의 URL 반환
+        return blob_client.url
+    except Exception as e:
+        print(f"An error occurred while uploading MP3 to Azure: {e}")
+        raise HTTPException(status_code=500, detail=f"An error occurred while uploading MP3 to Azure: {e}")

app/services/log_service.py

@@ -1,18 +1,22 @@
 # services/log_service.py
 from typing import List, Optional
+from sqlalchemy import null
 from sqlalchemy.ext.asyncio import AsyncSession
 from fastapi import HTTPException
 from app.crud.crud_log import (create_log as crud_create_log,
                           get_log as crud_get_log,
                           get_logs as crud_get_logs,
-                          update_log as crud_update_log,
                           delete_log as crud_delete_log)
-from app.schemas.LogDto import CreateLogDto
+from app.schemas.LogDto import CreateGreeTalkLogDto, CreateUserTalkLogDto
 from app.models.models import Log
 from sqlalchemy.future import select
 
-async def create_log_service(db: AsyncSession, log_dto: CreateLogDto) -> Log:
-    db_log = await crud_create_log(db, log_dto.gree_id, log_dto.log_type, log_dto.content)
+async def create_usertalk_log_service(db: AsyncSession, log_dto: CreateUserTalkLogDto) -> Log:
+    db_log = await crud_create_log(db, log_dto.gree_id, log_dto.log_type, log_dto.content, null)
+    return db_log
+
+async def create_greetalk_log_service(db: AsyncSession, log_dto: CreateGreeTalkLogDto) -> Log:
+    db_log = await crud_create_log(db, log_dto.gree_id, log_dto.log_type, log_dto.content, log_dto.voice_url)
     return db_log
 
 async def get_log_service(db: AsyncSession, log_id: int) -> Optional[Log]:

main.py

@@ -27,5 +27,5 @@
 app.include_router(api_router, prefix=settings.API_v1_STR)
 
 if __name__ == "__main__":
-    # init_db()
+    init_db()
     uvicorn.run("main:app", host="0.0.0.0", port=8000, reload=True, log_level="debug")