"""
LLM Provider Factory for Flare
"""
import os
from typing import Optional
from dotenv import load_dotenv

from llm_interface import LLMInterface, SparkLLM, GPT4oLLM
from config_provider import ConfigProvider
from utils import log

class LLMFactory:
    @staticmethod
    def create_provider() -> LLMInterface:
        """Create LLM provider based on configuration"""
        cfg = ConfigProvider.get()
        llm_config = cfg.global_config.llm_provider
        
        if not llm_config:
            raise ValueError("No LLM provider configured")
        
        provider_name = llm_config.name
        log(f"🏭 Creating LLM provider: {provider_name}")
        
        # Get provider definition
        provider_def = cfg.global_config.get_provider_config("llm", provider_name)
        if not provider_def:
            raise ValueError(f"Unknown LLM provider: {provider_name}")
        
        # Get API key
        api_key = LLMFactory._get_api_key(provider_name, llm_config.api_key)
        
        # Create provider based on name
        if provider_name == "spark":
            return LLMFactory._create_spark_provider(llm_config, api_key, provider_def)
        elif provider_name in ["gpt4o", "gpt4o-mini"]:
            return LLMFactory._create_gpt_provider(llm_config, api_key, provider_def)
        else:
            raise ValueError(f"Unsupported LLM provider: {provider_name}")
    
    @staticmethod
    def _create_spark_provider(llm_config, api_key: str, provider_def) -> SparkLLM:
        """Create Spark LLM provider"""
        if not llm_config.endpoint:
            raise ValueError("Spark endpoint is required")
        
        if not api_key:
            raise ValueError("Spark API token is required")
        
        # Extract work mode variant (for backward compatibility)
        provider_variant = "cloud"  # Default
        if os.getenv("SPACE_ID"):  # HuggingFace Space
            provider_variant = "hfcloud"
        
        log(f"🚀 Initializing SparkLLM: {llm_config.endpoint}")
        log(f"🔧 Provider variant: {provider_variant}")
        
        return SparkLLM(
            spark_endpoint=llm_config.endpoint,
            spark_token=api_key,
            provider_variant=provider_variant,
            settings=llm_config.settings
        )
    
    @staticmethod
    def _create_gpt_provider(llm_config, api_key: str, provider_def) -> GPT4oLLM:
        """Create GPT-4 provider"""
        if not api_key:
            raise ValueError("OpenAI API key is required")
        
        # Get model-specific settings
        settings = llm_config.settings or {}
        model = provider_def.name  # gpt4o or gpt4o-mini
        
        log(f"🤖 Initializing GPT4oLLM with model: {model}")
        
        return GPT4oLLM(
            api_key=api_key,
            model=model,
            settings=settings
        )
    
    @staticmethod
    def _get_api_key(provider_name: str, config_key: Optional[str]) -> Optional[str]:
        """Get API key from config or environment"""
        # First check config
        if config_key:
            if config_key.startswith("enc:"):
                # Decrypt if encrypted
                from encryption_utils import decrypt
                decrypted = decrypt(config_key)
                log(f"🔓 Using encrypted API key from config: ***{decrypted[-4:]}")
                return decrypted
            else:
                log(f"🔑 Using plain API key from config: ***{config_key[-4:]}")
                return config_key
        
        # Then check environment
        env_mappings = {
            "spark": "SPARK_TOKEN",
            "gpt4o": "OPENAI_API_KEY",
            "gpt4o-mini": "OPENAI_API_KEY"
        }
        
        env_var = env_mappings.get(provider_name)
        if env_var:
            # Check if we're in HuggingFace Space
            if os.getenv("SPACE_ID"):
                # HuggingFace mode - direct environment
                api_key = os.environ.get(env_var)
                if api_key:
                    log(f"🔑 Using API key from HuggingFace secrets: {env_var}")
                    return api_key
            else:
                # Local mode - use dotenv
                load_dotenv()
                api_key = os.getenv(env_var)
                if api_key:
                    log(f"🔑 Using API key from .env: {env_var}")
                    return api_key
        
        return None