from typing import Dict, Type
from src.core.settings import Settings
from src.libs.llm.azure_llm import AzureOpenAILLM
from src.libs.llm.base_llm import BaseLLM
from src.libs.llm.deepseek_llm import DeepSeekLLM
from src.libs.llm.ollama_llm import OllamaLLM
from src.libs.llm.openai_llm import OpenAILLM
class LLMFactory:
"""
Factory class for creating LLM instances based on configuration.
"""
_registry: Dict[str, Type] = {}
@staticmethod
def register(provider: str, llm_cls: Type) -> None:
LLMFactory._registry[str(provider).lower()] = llm_cls
@staticmethod
def create(settings: Settings) -> BaseLLM:
"""
Create an LLM instance based on the provided settings.
Args:
settings: Application settings containing LLM configuration
Returns:
An instance of a class implementing BaseLLM
Raises:
ValueError: If the configured provider is not supported or missing config.
"""
provider = settings.llm.provider.lower()
if provider in LLMFactory._registry:
return LLMFactory._registry[provider](settings.llm)
if provider == "openai":
if not settings.llm.api_key:
raise ValueError("OpenAI provider requires api_key")
return OpenAILLM(
api_key=settings.llm.api_key,
model=settings.llm.model,
base_url=settings.llm.base_url,
)
elif provider == "azure":
if not settings.llm.api_key:
raise ValueError("Azure provider requires api_key")
if not settings.llm.azure_endpoint:
raise ValueError("Azure provider requires azure_endpoint")
return AzureOpenAILLM(
api_key=settings.llm.api_key,
azure_endpoint=settings.llm.azure_endpoint,
model=settings.llm.model,
)
elif provider == "deepseek":
if not settings.llm.api_key:
raise ValueError("DeepSeek provider requires api_key")
# base_url is optional for DeepSeekLLM as it has a default
kwargs = {}
if settings.llm.base_url:
kwargs["base_url"] = settings.llm.base_url
return DeepSeekLLM(
api_key=settings.llm.api_key, model=settings.llm.model, **kwargs
)
elif provider == "ollama":
# base_url is optional (defaults to localhost), api_key is optional
kwargs = {}
if settings.llm.base_url:
kwargs["base_url"] = settings.llm.base_url
if settings.llm.api_key:
kwargs["api_key"] = settings.llm.api_key
return OllamaLLM(model=settings.llm.model, **kwargs)
else:
raise ValueError(f"Unknown LLM provider: {provider}")
@staticmethod
def create_vision(settings: Settings) -> BaseLLM:
"""
Create a Vision LLM instance based on the provided settings.
Args:
settings: Application settings containing Vision LLM configuration
Returns:
An instance of a class implementing BaseLLM
Raises:
ValueError: If the configured provider is not supported or missing config.
"""
provider = settings.vision_llm.provider.lower()
# Prioritize vision_llm specific settings, fallback to main LLM settings if needed
api_key = settings.vision_llm.api_key or settings.llm.api_key
base_url = settings.vision_llm.base_url or settings.llm.base_url
azure_endpoint = settings.vision_llm.azure_endpoint or settings.llm.azure_endpoint
if provider == "openai":
if not api_key:
raise ValueError("OpenAI provider requires api_key")
return OpenAILLM(
api_key=api_key, model=settings.vision_llm.model, base_url=base_url
)
elif provider == "azure":
if not api_key:
raise ValueError("Azure provider requires api_key")
if not azure_endpoint:
raise ValueError("Azure provider requires azure_endpoint")
return AzureOpenAILLM(
api_key=api_key,
azure_endpoint=azure_endpoint,
model=settings.vision_llm.model,
)
elif provider == "ollama":
if base_url:
return OllamaLLM(model=settings.vision_llm.model, base_url=base_url)
return OllamaLLM(model=settings.vision_llm.model)
else:
raise ValueError(f"Unknown Vision LLM provider: {provider}")
