Browser-Use MCP Server

"""LLM provider factory using browser-use native providers.""" from typing import TYPE_CHECKING # Import available chat models from browser-use from browser_use import ( ChatAnthropic, ChatAzureOpenAI, ChatBrowserUse, ChatGoogle, ChatGroq, ChatOllama, ChatOpenAI, ChatVercel, ) # These are available via direct import but not in __all__ from browser_use.llm.aws.chat_bedrock import ChatAWSBedrock from browser_use.llm.cerebras.chat import ChatCerebras from browser_use.llm.deepseek.chat import ChatDeepSeek from browser_use.llm.openrouter.chat import ChatOpenRouter from .config import NO_KEY_PROVIDERS, STANDARD_ENV_VAR_NAMES from .exceptions import LLMProviderError if TYPE_CHECKING: from browser_use.llm.base import BaseChatModel def get_llm( provider: str, model: str, api_key: str | None = None, base_url: str | None = None, **kwargs, ) -> "BaseChatModel": """Create LLM instance using browser-use native providers. Supports 12 providers: - openai: OpenAI GPT models - anthropic: Claude models - google: Gemini models - azure_openai: Azure-hosted OpenAI models - groq: Groq-hosted models - deepseek: DeepSeek models - cerebras: Cerebras models - ollama: Local Ollama models (no API key required) - bedrock: AWS Bedrock models (uses AWS credentials) - browser_use: Browser Use Cloud API - openrouter: OpenRouter API - vercel: Vercel AI Gateway Args: provider: LLM provider name model: Model name/identifier api_key: API key for the provider (not required for ollama/bedrock) base_url: Custom base URL for OpenAI-compatible APIs **kwargs: Provider-specific options: - azure_endpoint: Azure OpenAI endpoint URL - azure_api_version: Azure OpenAI API version (default: 2024-02-01) - aws_region: AWS region for Bedrock Returns: Configured BaseChatModel instance Raises: LLMProviderError: If provider is unsupported or API key is missing """ # Check if API key is required requires_api_key = provider not in NO_KEY_PROVIDERS and not base_url if requires_api_key and not api_key: standard_var = STANDARD_ENV_VAR_NAMES.get(provider, "API key") raise LLMProviderError(f"API key required for provider '{provider}'. Set {standard_var} or MCP_LLM_API_KEY environment variable.") try: match provider: case "openai": return ChatOpenAI(model=model, api_key=api_key, base_url=base_url) case "anthropic": return ChatAnthropic(model=model, api_key=api_key) case "google": return ChatGoogle(model=model, api_key=api_key) case "azure_openai": azure_endpoint = kwargs.get("azure_endpoint") azure_api_version = kwargs.get("azure_api_version", "2024-02-01") if not azure_endpoint: raise LLMProviderError("Azure OpenAI requires AZURE_OPENAI_ENDPOINT or MCP_LLM_AZURE_ENDPOINT to be set.") return ChatAzureOpenAI( model=model, api_key=api_key, azure_endpoint=azure_endpoint, api_version=azure_api_version, ) case "groq": return ChatGroq(model=model, api_key=api_key) case "deepseek": return ChatDeepSeek(model=model, api_key=api_key) case "cerebras": return ChatCerebras(model=model, api_key=api_key) case "ollama": return ChatOllama(model=model, host=base_url) case "bedrock": aws_region = kwargs.get("aws_region") return ChatAWSBedrock(model=model, aws_region=aws_region) case "browser_use": return ChatBrowserUse(model=model, api_key=api_key) case "openrouter": return ChatOpenRouter(model=model, api_key=api_key) case "vercel": return ChatVercel(model=model, api_key=api_key) case _: raise LLMProviderError(f"Unsupported provider: {provider}") except LLMProviderError: raise except Exception as e: raise LLMProviderError(f"Failed to initialize {provider} LLM: {e}") from e