OpenRouter MCP Server

""" Configuration management for OpenRouter MCP Server """ import os from typing import Dict, Any, Optional from dotenv import load_dotenv # Load environment variables load_dotenv() # Version and metadata VERSION = "1.0.0" MCP_VERSION = "1.0" # OpenRouter API Configuration OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY") OPENROUTER_BASE_URL = os.getenv("OPENROUTER_BASE_URL", "https://openrouter.ai/api/v1") # Default model settings DEFAULT_MODEL = os.getenv("DEFAULT_MODEL", "moonshotai/kimi-k2-thinking") DEFAULT_TEMPERATURE = float(os.getenv("DEFAULT_TEMPERATURE", "0.7")) DEFAULT_MAX_TOKENS = int(os.getenv("DEFAULT_MAX_TOKENS", "8192")) DEFAULT_MAX_REASONING_TOKENS = int( os.getenv("DEFAULT_MAX_REASONING_TOKENS", "16384") ) # Max thinking/reasoning tokens # Tool configuration ENABLE_WEB_SEARCH = os.getenv("ENABLE_WEB_SEARCH", "true").lower() == "true" FORCE_INTERNET_SEARCH = os.getenv("FORCE_INTERNET_SEARCH", "true").lower() == "true" # Logging configuration LOG_LEVEL = os.getenv("LOG_LEVEL", "INFO") LOG_FILE = os.getenv("LOG_FILE", "openrouter_mcp.log") # Rate limiting RATE_LIMIT_REQUESTS_PER_MINUTE = int(os.getenv("RATE_LIMIT_REQUESTS_PER_MINUTE", "60")) # MCP Transport limits MAX_MESSAGE_SIZE = int(os.getenv("MAX_MESSAGE_SIZE", "10485760")) # 10MB MAX_CONCURRENT_REQUESTS = int(os.getenv("MAX_CONCURRENT_REQUESTS", "10")) # OpenRouter-specific model configurations PREFERRED_MODELS = { "gemini-2.5-pro": "google/gemini-2.5-pro", "gemini-pro": "google/gemini-2.5-pro", "deepseek-r1": "deepseek/deepseek-r1-0528", "deepseek": "deepseek/deepseek-r1-0528", "deepseek-v3.1": "deepseek/deepseek-chat-v3.1", "deepseek-chat-v3": "deepseek/deepseek-chat-v3.1", "kimi-k2": "moonshotai/kimi-k2-0905", "kimi": "moonshotai/kimi-k2-0905", "grok-4": "x-ai/grok-code-fast-1", "grok": "x-ai/grok-code-fast-1", "qwen3-max": "qwen/qwen3-max", "qwen-max": "qwen/qwen3-max", "qwen3-coder-plus": "qwen/qwen3-coder-plus", "qwen3-coder": "qwen/qwen3-coder-plus", "qwen-coder": "qwen/qwen3-coder-plus", "glm-4.6": "z-ai/glm-4.6", "glm": "z-ai/glm-4.6", "gpt-5": "openai/gpt-5", "openai-gpt-5": "openai/gpt-5", } # Model capabilities configuration MODEL_CAPABILITIES = { "vision": ["google/gemini-2.5-pro", "openai/gpt-5"], "function_calling": ["google/gemini-2.5-pro", "openai/gpt-5"], "large_context": [ "deepseek/deepseek-r1-0528", "deepseek/deepseek-chat-v3.1", "google/gemini-2.5-pro", "moonshotai/kimi-k2-0905", "x-ai/grok-code-fast-1", "qwen/qwen3-max", "qwen/qwen3-coder-plus", "z-ai/glm-4.6", "openai/gpt-5", ], "internet_access": ["google/gemini-2.5-pro"], } def get_config() -> Dict[str, Any]: """Get current configuration as dictionary""" return { "version": VERSION, "mcp_version": MCP_VERSION, "openrouter": { "api_key": bool(OPENROUTER_API_KEY), # Don't expose actual key "base_url": OPENROUTER_BASE_URL, }, "defaults": { "model": DEFAULT_MODEL, "temperature": DEFAULT_TEMPERATURE, "max_tokens": DEFAULT_MAX_TOKENS, "max_reasoning_tokens": DEFAULT_MAX_REASONING_TOKENS, }, "tools": { "web_search": ENABLE_WEB_SEARCH, "force_internet_search": FORCE_INTERNET_SEARCH, }, "logging": { "level": LOG_LEVEL, "file": LOG_FILE, }, "rate_limits": { "requests_per_minute": RATE_LIMIT_REQUESTS_PER_MINUTE, }, "transport": { "max_message_size": MAX_MESSAGE_SIZE, "max_concurrent_requests": MAX_CONCURRENT_REQUESTS, }, } def validate_config() -> bool: """Validate configuration and return True if valid""" if not OPENROUTER_API_KEY: print("ERROR: OPENROUTER_API_KEY is required") return False if DEFAULT_TEMPERATURE < 0 or DEFAULT_TEMPERATURE > 2: print("ERROR: DEFAULT_TEMPERATURE must be between 0 and 2") return False if DEFAULT_MAX_TOKENS < 1: print("ERROR: DEFAULT_MAX_TOKENS must be positive") return False return True def get_model_alias(model_name: str, user_prompt: str = "") -> str: """Get the actual OpenRouter model name using intelligent LLM-based selection""" if not model_name: return DEFAULT_MODEL # Direct alias match first if model_name in PREFERRED_MODELS: return PREFERRED_MODELS[model_name] # If it's already a full OpenRouter model name, return as-is if "/" in model_name and model_name not in PREFERRED_MODELS: return model_name # Use LLM intelligence to determine the best model based on user query return _intelligent_model_selection(model_name, user_prompt) def _intelligent_model_selection(model_request: str, user_prompt: str = "") -> str: """Use LLM intelligence to select the best model based on context""" # Model capabilities for intelligent selection model_info = { "gemini-2.5-pro": { "model": "google/gemini-2.5-pro", "strengths": "vision, web search, general reasoning, large context (1M+ tokens)", "best_for": "image analysis, current information, research, general tasks", }, "deepseek-r1": { "model": "deepseek/deepseek-r1-0528", "strengths": "advanced reasoning, logical analysis, problem solving", "best_for": "complex reasoning, mathematical problems, logical analysis", }, "deepseek-v3.1": { "model": "deepseek/deepseek-chat-v3.1", "strengths": "latest version with 163K context, advanced chat capabilities", "best_for": "general chat, latest features, large context tasks", }, "kimi-k2": { "model": "moonshotai/kimi-k2-0905", "strengths": "advanced reasoning, programming, large context", "best_for": "programming tasks, code analysis, advanced reasoning", }, "grok-4": { "model": "x-ai/grok-code-fast-1", "strengths": "fast code generation, programming tasks, technical solutions", "best_for": "code generation, debugging, programming assistance, technical tasks", }, "qwen3-max": { "model": "qwen/qwen3-max", "strengths": "large context (128K), general reasoning, multilingual", "best_for": "large document analysis, general tasks, multilingual content", }, "qwen3-coder-plus": { "model": "qwen/qwen3-coder-plus", "strengths": "coding, programming, technical tasks (32K context)", "best_for": "code generation, debugging, programming assistance", }, "glm-4.6": { "model": "z-ai/glm-4.6", "strengths": "balanced performance, general tasks, good default choice", "best_for": "general purpose tasks, balanced performance", }, "gpt-5": { "model": "openai/gpt-5", "strengths": "flagship model with 400K context, latest capabilities", "best_for": "cutting-edge performance, large context tasks, latest features", }, } # Simple intelligent matching based on request context request_lower = model_request.lower().strip() prompt_lower = user_prompt.lower() if user_prompt else "" # Direct name matching with intelligence if "gemini" in request_lower or "google" in request_lower: return model_info["gemini-2.5-pro"]["model"] elif "deepseek" in request_lower: # Check for version preference if any(word in request_lower for word in ["v3.1", "v3", "chat", "latest"]): return model_info["deepseek-v3.1"]["model"] else: return model_info["deepseek-r1"]["model"] elif "kimi" in request_lower or "moonshot" in request_lower: return model_info["kimi-k2"]["model"] elif "grok" in request_lower or "x-ai" in request_lower or "xai" in request_lower: return model_info["grok-4"]["model"] elif "glm" in request_lower or "z-ai" in request_lower: return model_info["glm-4.6"]["model"] elif ( "gpt-5" in request_lower or "gpt5" in request_lower or "openai" in request_lower ): return model_info["gpt-5"]["model"] # For qwen, use context to determine which variant elif "qwen" in request_lower: # Analyze user prompt to determine best qwen variant if any( word in prompt_lower for word in [ "code", "programming", "debug", "function", "script", "development", ] ): return model_info["qwen3-coder-plus"]["model"] else: return model_info["qwen3-max"]["model"] # If no match found, return the request as-is (assume it's a full model name) return model_request def list_available_aliases() -> dict: """List all available model aliases and their mappings""" return PREFERRED_MODELS.copy() def suggest_model_alias(partial_name: str) -> list: """Suggest model aliases based on partial input""" if not partial_name: return [] partial_lower = partial_name.lower() suggestions = [] for alias in PREFERRED_MODELS.keys(): if partial_lower in alias.lower(): suggestions.append(alias) return sorted(suggestions) def has_capability(model_name: str, capability: str) -> bool: """Check if a model has a specific capability""" actual_model = get_model_alias(model_name) return actual_model in MODEL_CAPABILITIES.get(capability, []) def should_force_internet_search(model_name: str) -> bool: """Check if we should force internet search for this model""" if not FORCE_INTERNET_SEARCH: return False return has_capability(model_name, "internet_access")