feat: AI 서비스 MLX 듀얼 백엔드 및 모델 최적화

- MLX(맥미니 27B) 우선 → Ollama(조립컴 9B) fallback 구조
- pydantic-settings 기반 config 전환
- health check에 MLX 상태 추가
- 텍스트 모델 qwen3:8b → qwen3.5:9b-q8_0 변경

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

ai-service/middlewares/auth.py

@@ -0,0 +1,24 @@
+from fastapi import Request, HTTPException, status
+from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
+from jose import jwt, JWTError, ExpiredSignatureError
+from config import settings
+
+security = HTTPBearer(auto_error=False)
+
+
+async def verify_token(request: Request) -> dict:
+    """JWT 토큰 검증. SSO 서비스와 동일한 시크릿 사용."""
+    auth: HTTPAuthorizationCredentials = await security(request)
+    if not auth:
+        raise HTTPException(status_code=status.HTTP_401_UNAUTHORIZED, detail="Authorization 헤더가 필요합니다")
+
+    if not settings.SECRET_KEY:
+        raise HTTPException(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail="서버 인증 설정 오류")
+
+    try:
+        payload = jwt.decode(auth.credentials, settings.SECRET_KEY, algorithms=[settings.ALGORITHM])
+        return payload
+    except ExpiredSignatureError:
+        raise HTTPException(status_code=status.HTTP_401_UNAUTHORIZED, detail="토큰이 만료되었습니다")
+    except JWTError:
+        raise HTTPException(status_code=status.HTTP_401_UNAUTHORIZED, detail="유효하지 않은 토큰입니다")