core/config.py

import os
from pathlib import Path
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_openai import ChatOpenAI
from dotenv import load_dotenv
import logging
from logging.handlers import RotatingFileHandler
from pydantic_settings import BaseSettings, SettingsConfigDict

# Initialize environment
load_dotenv()


# --- Settings (simple, in-file) ---
class Settings(BaseSettings):
    model_config = SettingsConfigDict(env_file='.env', env_file_encoding='utf-8', extra='ignore')

    OPENAI_API_KEY: str
    OPENAI_BASE_URL: str | None = None
    
    # LlamaParse configuration for advanced PDF parsing
    LLAMA_CLOUD_API_KEY: str | None = None
    LLAMA_PREMIUM_MODE: bool = False  # Set to True for GPT-4o parsing (costs more)

    LOG_LEVEL: str = os.getenv("LOG_LEVEL", "INFO")
    DATA_DIR: str = os.getenv("DATA_DIR", "")
    LOG_DIR: str = os.getenv("LOG_DIR", "")


settings = Settings()


# --- File Path Configuration (Cross-platform compatible) ---
PROJECT_ROOT = Path(__file__).parent.parent.absolute()
DATA_DIR = Path(settings.DATA_DIR or (PROJECT_ROOT / "data"))
NEW_DATA = DATA_DIR / "new_data"
PROCESSED_DATA = DATA_DIR / "processed_data"
CHUNKS_PATH = DATA_DIR / "chunks.pkl"
VECTOR_STORE_DIR = DATA_DIR / "vector_store"
DATA_DIR.mkdir(parents=True, exist_ok=True)
NEW_DATA.mkdir(parents=True, exist_ok=True)
PROCESSED_DATA.mkdir(parents=True, exist_ok=True)
VECTOR_STORE_DIR.mkdir(parents=True, exist_ok=True)

# Setup logging
LOG_DIR = Path(settings.LOG_DIR or (Path(__file__).parent.parent / "logs"))
LOG_DIR.mkdir(parents=True, exist_ok=True)

LOG_FILE = LOG_DIR / "app.log"

# Configure application logger (avoid duplicate handlers)
LOG_LEVEL = settings.LOG_LEVEL.upper()
logger = logging.getLogger("AgenticMedicalRAG")  # centralized logger
logger.setLevel(LOG_LEVEL)
logger.propagate = False
if not logger.handlers:
    formatter = logging.Formatter(
        fmt="%(asctime)s - %(name)s - %(levelname)s - %(message)s"
    )
    file_handler = RotatingFileHandler(
        LOG_FILE,
        maxBytes=1000000,
        backupCount=3,
        encoding="utf-8"
    )
    file_handler.setFormatter(formatter)
    stream_handler = logging.StreamHandler()
    stream_handler.setFormatter(formatter)
    logger.addHandler(file_handler)
    logger.addHandler(stream_handler)



# --- LLM Configuration with lazy loading ---
_llm = None

def get_llm():
    """Get LLM with lazy loading for faster startup"""
    global _llm
    if _llm is None:
        logger.info("Initializing LLM (first time)...")
        openai_key = settings.OPENAI_API_KEY
        
        if not openai_key:
            logger.error("OPENAI_API_KEY not found in environment variables")
            raise ValueError("OpenAI API key is required. Please set OPENAI_API_KEY environment variable.")
        
        try:
            _llm = ChatOpenAI(
                model="gpt-4o",
                api_key=openai_key,
                base_url=settings.OPENAI_BASE_URL,
                temperature=0.0,
                max_tokens=2048,
                request_timeout=30,  # Increased timeout for stability
                max_retries=2,
                streaming=True,
            )
            logger.info("LLM initialized successfully")
        except Exception as e:
            logger.error(f"Failed to initialize LLM: {e}")
            raise
    return _llm

def create_llm():
    """Create LLM with proper error handling and fallbacks"""
    return get_llm()

# Lazy loading - only initialize when actually needed
LLM = None  # Will be loaded on first use

# --- Embedding Model Configuration with lazy loading ---
_embedding_model = None

def get_embedding_model():
    """Get embedding model with lazy loading for faster startup"""
    global _embedding_model
    if _embedding_model is None:
        logger.info("Loading embedding model (first time)...")
        try:
            _embedding_model = HuggingFaceEmbeddings(
                model_name="abhinand/MedEmbed-base-v0.1",
                model_kwargs={'device': 'cpu'},
                encode_kwargs={'normalize_embeddings': True}
            )
            logger.info("Embedding model loaded successfully")
        except Exception as e:
            logger.error(f"Failed to load embedding model: {e}")
            raise ValueError("Failed to load embedding model")
    return _embedding_model

# For backward compatibility
def create_embedding_model():
    """Create embedding model with proper error handling"""
    return get_embedding_model()

# Lazy loading - only load when actually needed
EMBEDDING_MODEL = None  # Will be loaded on first use

# Configuration validation
def validate_config():
    """Validate all required configurations"""
    required_env_vars = ["OPENAI_API_KEY"]
    missing_vars = [var for var in required_env_vars if not getattr(settings, var, None)]
    
    if missing_vars:
        raise ValueError(f"Missing required environment variables: {missing_vars}")
    
    logger.info("Configuration validation completed")

# Run validation on import
try:
    validate_config()
except Exception as e:
    logger.error(f"Configuration validation failed: {e}")
    raise e