agent-mcp 0.1.3__py3-none-any.whl → 0.1.5__py3-none-any.whl

agent_mcp/pydantic_ai_mcp_adapter.py

@@ -0,0 +1,628 @@
+"""
+Pydantic AI MCP Integration
+FastAPI-like agent framework with native MCP support
+
+This module provides integration between Pydantic AI agents and the Model Context Protocol,
+offering type-safe, production-ready agent capabilities.
+"""
+
+import asyncio
+import json
+import uuid
+from typing import Dict, Any, List, Optional, Callable, Union, Type
+from dataclasses import dataclass, asdict
+from datetime import datetime, timezone
+import logging
+
+logger = logging.getLogger(__name__)
+
+# Try to import Pydantic AI
+try:
+    from pydantic_ai import Agent, RunContext
+    from pydantic_ai.models import OpenAIModel
+    # from pydantic_ai.tools import Tool  # Not available in all versions
+    from pydantic_ai.mcp import MCPServerTool, FastMCPToolset
+    PYDANTIC_AI_AVAILABLE = True
+except ImportError:
+    PYDANTIC_AI_AVAILABLE = False
+    Agent = None
+    RunContext = None
+    logger.warning("Pydantic AI not available. Install with: pip install pydantic-ai")
+
+from .mcp_transport import HTTPTransport
+
+@dataclass
+class PydanticAIAgentConfig:
+    """Configuration for Pydantic AI agent integration"""
+    agent_id: str
+    name: str
+    description: str
+    model: str = "openai:gpt-4o"
+    model_type: str = "openai"  # "openai", "anthropic", "gemini"
+    api_key: str = None
+    mcp_server_url: str = "https://mcp-server-ixlfhxquwq-ew.a.run.app"
+    enable_type_validation: bool = True
+    enable_retry: bool = True
+    max_retries: int = 3
+    timeout: int = 30
+    debug_mode: bool = False
+
+class PydanticAIMCPAgent:
+    """Pydantic AI Agent with MCP capabilities"""
+    
+    def __init__(
+        self,
+        config: PydanticAIAgentConfig,
+        transport: HTTPTransport = None
+    ):
+        if not PYDANTIC_AI_AVAILABLE:
+            raise ImportError("Pydantic AI is not installed. Install with: pip install pydantic-ai")
+            
+        self.config = config
+        self.transport = transport or HTTPTransport.from_url(config.mcp_server_url)
+        self.mcp_id = config.agent_id
+        self.mcp_version = "0.1.0"
+        self.mcp_tools = {}
+        
+        # Initialize Pydantic AI agent
+        self.agent = None
+        self._setup_pydantic_agent()
+        
+        # Register MCP tools
+        self._register_default_mcp_tools()
+    
+    def _setup_pydantic_agent(self):
+        """Setup the Pydantic AI agent"""
+        try:
+            # Configure model based on type
+            if self.config.model_type == "openai":
+                model = OpenAIModel(
+                    model_name=self.config.model.replace("openai:", ""),
+                    api_key=self.config.api_key
+                )
+            else:
+                # For other model types, you'd need to configure appropriately
+                model = OpenAIModel(
+                    model_name="gpt-4o",
+                    api_key=self.config.api_key
+                )
+            
+            # Create the Pydantic AI agent
+            self.agent = Agent(
+                name=self.config.name,
+                model=model,
+                system_prompt=self.config.description,
+                retries=self.config.max_retries if self.config.enable_retry else 0
+            )
+            
+            logger.info(f"Created Pydantic AI agent: {self.config.name}")
+            
+        except Exception as e:
+            logger.error(f"Error setting up Pydantic AI agent: {e}")
+    
+    def _register_default_mcp_tools(self):
+        """Register default MCP tools for Pydantic AI agent"""
+        
+        # Pydantic AI specific tools
+        self._register_pydantic_tools_as_mcp()
+        
+        # MCP integration tools
+        self._register_mcp_integration_tools()
+        
+        # Agent management tools
+        self._register_agent_management_tools()
+    
+    def _register_pydantic_tools_as_mcp(self):
+        """Register Pydantic AI tools as MCP tools"""
+        
+        # Get tools from the agent if available
+        if hasattr(self.agent, 'tools') and self.agent.tools:
+            for tool_name, tool in self.agent.tools.items():
+                self._register_tool_as_mcp(tool_name, tool)
+    
+    def _register_tool_as_mcp(self, tool_name: str, tool: Tool):
+        """Register a Pydantic AI tool as MCP tool"""
+        
+        async def pydantic_tool_wrapper(**kwargs):
+            """Wrapper to call Pydantic AI tool through MCP"""
+            try:
+                # Execute the tool with type validation
+                if self.config.enable_type_validation:
+                    # Pydantic AI handles type validation automatically
+                    result = await tool(**kwargs)
+                else:
+                    result = await tool(**kwargs)
+                
+                return {
+                    "status": "success",
+                    "result": result,
+                    "tool_name": tool_name,
+                    "framework": "pydantic_ai"
+                }
+            except Exception as e:
+                logger.error(f"Error in Pydantic AI tool {tool_name}: {e}")
+                return {
+                    "status": "error",
+                    "message": str(e),
+                    "tool_name": tool_name
+                }
+        
+        # Extract tool metadata
+        tool_info = {
+            "name": tool_name,
+            "description": getattr(tool, 'description', f"Pydantic AI tool {tool_name}"),
+            "parameters": self._extract_pydantic_tool_parameters(tool),
+            "function": pydantic_tool_wrapper
+        }
+        
+        self.mcp_tools[tool_name] = tool_info
+    
+    def _extract_pydantic_tool_parameters(self, tool: Tool) -> List[Dict[str, Any]]:
+        """Extract parameters from Pydantic AI tool"""
+        parameters = []
+        
+        try:
+            # Pydantic AI tools have function signatures we can inspect
+            import inspect
+            sig = inspect.signature(tool)
+            
+            for param_name, param in sig.parameters.items():
+                if param_name == 'ctx':  # Skip RunContext parameter
+                    continue
+                    
+                parameter = {
+                    "name": param_name,
+                    "description": f"Parameter {param_name}",
+                    "type": "string",  # Default to string
+                    "required": param.default == inspect.Parameter.empty
+                }
+                
+                # Try to determine type from annotation
+                if param.annotation != inspect.Parameter.empty:
+                    type_str = str(param.annotation)
+                    if "int" in type_str.lower():
+                        parameter["type"] = "number"
+                    elif "bool" in type_str.lower():
+                        parameter["type"] = "boolean"
+                    elif "list" in type_str.lower():
+                        parameter["type"] = "array"
+                    elif "dict" in type_str.lower():
+                        parameter["type"] = "object"
+                
+                parameters.append(parameter)
+                
+        except Exception as e:
+            logger.error(f"Error extracting parameters from Pydantic AI tool: {e}")
+        
+        return parameters
+    
+    def _register_mcp_integration_tools(self):
+        """Register MCP-specific integration tools"""
+        
+        async def connect_to_mcp_server(server_url: str, tools: List[str] = None) -> Dict[str, Any]:
+            """Connect to an external MCP server and use its tools"""
+            try:
+                if FastMCPToolset and MCPServerTool:
+                    # Use Pydantic AI's built-in MCP support
+                    mcp_toolset = FastMCPToolset(server_url)
+                    
+                    # Add specific tools if requested
+                    if tools:
+                        available_tools = []
+                        for tool_name in tools:
+                            try:
+                                tool = MCPServerTool(server_url, tool_name)
+                                available_tools.append(tool)
+                            except Exception as e:
+                                logger.warning(f"Could not load MCP tool {tool_name}: {e}")
+                        
+                        if available_tools:
+                            # Add tools to agent
+                            self.agent.tools.update({
+                                f"mcp_{tool.name}": tool for tool in available_tools
+                            })
+                    
+                    return {
+                        "status": "success",
+                        "message": f"Connected to MCP server at {server_url}",
+                        "tools_loaded": len(available_tools) if 'available_tools' in locals() else 0
+                    }
+                else:
+                    # Fallback: manual MCP connection
+                    import aiohttp
+                    async with aiohttp.ClientSession() as session:
+                        async with session.get(f"{server_url}/tools") as response:
+                            if response.status == 200:
+                                tools_data = await response.json()
+                                return {
+                                    "status": "success",
+                                    "server_url": server_url,
+                                    "available_tools": tools_data.get("tools", [])
+                                }
+                            else:
+                                return {
+                                    "status": "error",
+                                    "message": f"Failed to connect to MCP server: HTTP {response.status}"
+                                }
+            except Exception as e:
+                logger.error(f"Error connecting to MCP server {server_url}: {e}")
+                return {
+                    "status": "error",
+                    "message": str(e),
+                    "server_url": server_url
+                }
+        
+        self.mcp_tools["connect_to_mcp_server"] = {
+            "name": "connect_to_mcp_server",
+            "description": "Connect to an external MCP server and import its tools",
+            "parameters": [
+                {
+                    "name": "server_url",
+                    "description": "URL of the MCP server to connect to",
+                    "type": "string",
+                    "required": True
+                },
+                {
+                    "name": "tools",
+                    "description": "List of specific tools to import (optional)",
+                    "type": "array",
+                    "required": False
+                }
+            ],
+            "function": connect_to_mcp_server
+        }
+    
+    def _register_agent_management_tools(self):
+        """Register agent management and monitoring tools"""
+        
+        async def get_agent_info() -> Dict[str, Any]:
+            """Get information about this Pydantic AI agent"""
+            return {
+                "status": "success",
+                "agent_info": {
+                    "id": self.mcp_id,
+                    "name": self.config.name,
+                    "description": self.config.description,
+                    "framework": "Pydantic AI",
+                    "model": self.config.model,
+                    "type_validation": self.config.enable_type_validation,
+                    "retry_enabled": self.config.enable_retry,
+                    "max_retries": self.config.max_retries,
+                    "mcp_tools": list(self.mcp_tools.keys())
+                }
+            }
+        
+        self.mcp_tools["get_agent_info"] = {
+            "name": "get_agent_info",
+            "description": "Get detailed information about this Pydantic AI agent",
+            "parameters": [],
+            "function": get_agent_info
+        }
+        
+        async def update_agent_config(
+            new_description: str = None,
+            enable_type_validation: bool = None,
+            max_retries: int = None
+        ) -> Dict[str, Any]:
+            """Update agent configuration dynamically"""
+            try:
+                if new_description:
+                    self.config.description = new_description
+                    # Update agent system prompt
+                    self.agent.system_prompt = new_description
+                
+                if enable_type_validation is not None:
+                    self.config.enable_type_validation = enable_type_validation
+                
+                if max_retries is not None:
+                    self.config.max_retries = max_retries
+                    self.agent.retries = max_retries
+                
+                return {
+                    "status": "success",
+                    "message": "Agent configuration updated",
+                    "config": {
+                        "description": self.config.description,
+                        "type_validation": self.config.enable_type_validation,
+                        "max_retries": self.config.max_retries
+                    }
+                }
+            except Exception as e:
+                logger.error(f"Error updating agent config: {e}")
+                return {
+                    "status": "error",
+                    "message": str(e)
+                }
+        
+        self.mcp_tools["update_agent_config"] = {
+            "name": "update_agent_config",
+            "description": "Update the configuration of this agent",
+            "parameters": [
+                {
+                    "name": "new_description",
+                    "description": "New system description",
+                    "type": "string",
+                    "required": False
+                },
+                {
+                    "name": "enable_type_validation",
+                    "description": "Enable Pydantic type validation",
+                    "type": "boolean",
+                    "required": False
+                },
+                {
+                    "name": "max_retries",
+                    "description": "Maximum number of retries",
+                    "type": "number",
+                    "required": False
+                }
+            ],
+            "function": update_agent_config
+        }
+    
+    def add_custom_tool(
+        self,
+        name: str,
+        description: str,
+        func: Callable,
+        parameters: List[Dict[str, Any]] = None
+    ):
+        """Add a custom tool to the Pydantic AI agent"""
+        
+        # Register with Pydantic AI agent
+        if self.agent:
+            self.agent.tools[name] = func
+        
+        # Register as MCP tool
+        async def custom_tool_wrapper(**kwargs):
+            """Wrapper for custom tool"""
+            try:
+                result = await func(**kwargs)
+                return {
+                    "status": "success",
+                    "result": result,
+                    "tool_name": name,
+                    "type": "custom"
+                }
+            except Exception as e:
+                logger.error(f"Error in custom tool {name}: {e}")
+                return {
+                    "status": "error",
+                    "message": str(e),
+                    "tool_name": name
+                }
+        
+        self.mcp_tools[name] = {
+            "name": name,
+            "description": description,
+            "parameters": parameters or [],
+            "function": custom_tool_wrapper
+        }
+    
+    async def run_agent(
+        self,
+        message: str,
+        context: Dict[str, Any] = None
+    ) -> Dict[str, Any]:
+        """Run the Pydantic AI agent with a message"""
+        try:
+            # Create run context if provided
+            run_context = None
+            if context:
+                run_context = RunContext(
+                    deps=context
+                )
+            
+            # Run the agent
+            result = await self.agent.run(message, ctx=run_context)
+            
+            return {
+                "status": "success",
+                "result": result.data if hasattr(result, 'data') else str(result),
+                "message": message,
+                "context_used": context is not None,
+                "agent_id": self.mcp_id
+            }
+            
+        except Exception as e:
+            logger.error(f"Error running Pydantic AI agent: {e}")
+            return {
+                "status": "error",
+                "message": str(e),
+                "message_sent": message
+            }
+    
+    async def register_with_mcp_server(self) -> Dict[str, Any]:
+        """Register this Pydantic AI agent with MCP server"""
+        registration_data = {
+            "agent_id": self.mcp_id,
+            "info": {
+                "name": self.config.name,
+                "description": self.config.description,
+                "framework": "Pydantic AI",
+                "model": self.config.model,
+                "capabilities": {
+                    "type_validation": self.config.enable_type_validation,
+                    "retry_mechanism": self.config.enable_retry,
+                    "max_retries": self.config.max_retries,
+                    "tools": list(self.mcp_tools.keys())
+                },
+                "version": self.mcp_version,
+                "features": {
+                    "type_safe": True,
+                    "fastapi_style": True,
+                    "production_ready": True,
+                    "debug_mode": self.config.debug_mode
+                }
+            }
+        }
+        
+        return await self.transport.register_agent(self)
+    
+    async def execute_mcp_tool(self, tool_name: str, **kwargs) -> Dict[str, Any]:
+        """Execute an MCP tool"""
+        if tool_name in self.mcp_tools:
+            tool_func = self.mcp_tools[tool_name]["function"]
+            return await tool_func(**kwargs)
+        else:
+            return {
+                "status": "error",
+                "message": f"Tool {tool_name} not found",
+                "available_tools": list(self.mcp_tools.keys())
+            }
+    
+    def get_mcp_tool_info(self) -> Dict[str, Any]:
+        """Get information about all available MCP tools"""
+        return {
+            "agent_id": self.mcp_id,
+            "framework": "Pydantic AI",
+            "model": self.config.model,
+            "tools": [
+                {
+                    "name": tool_info["name"],
+                    "description": tool_info["description"],
+                    "parameters": tool_info["parameters"]
+                }
+                for tool_info in self.mcp_tools.values()
+            ],
+            "total_tools": len(self.mcp_tools)
+        }
+
+class PydanticAIAgentBridge:
+    """Bridge for Pydantic AI agent integration with MCP"""
+    
+    def __init__(self, mcp_server_url: str = "https://mcp-server-ixlfhxquwq-ew.a.run.app"):
+        self.mcp_server_url = mcp_server_url
+        self.agents = {}
+    
+    async def create_agent(
+        self,
+        agent_id: str,
+        name: str,
+        description: str,
+        model: str = "openai:gpt-4o",
+        api_key: str = None,
+        **kwargs
+    ) -> PydanticAIMCPAgent:
+        """Create and register a Pydantic AI agent"""
+        
+        config = PydanticAIAgentConfig(
+            agent_id=agent_id,
+            name=name,
+            description=description,
+            model=model,
+            api_key=api_key,
+            mcp_server_url=self.mcp_server_url,
+            **kwargs
+        )
+        
+        agent = PydanticAIMCPAgent(config)
+        
+        # Register with MCP server
+        registration_result = await agent.register_with_mcp_server()
+        
+        if registration_result.get("status") == "registered":
+            logger.info(f"Pydantic AI agent {agent_id} registered with MCP server")
+            self.agents[agent_id] = agent
+        else:
+            logger.error(f"Failed to register Pydantic AI agent {agent_id}: {registration_result}")
+        
+        return agent
+    
+    async def create_multi_agent_system(
+        self,
+        agents_config: List[Dict[str, Any]]
+    ) -> Dict[str, PydanticAIMCPAgent]:
+        """Create a multi-agent system with Pydantic AI"""
+        
+        agents = {}
+        
+        for config in agents_config:
+            agent = await self.create_agent(**config)
+            agents[config.get("agent_id", agent.mcp_id)] = agent
+        
+        logger.info(f"Created Pydantic AI multi-agent system with {len(agents)} agents")
+        return agents
+    
+    async def setup_agent_collaboration(
+        self,
+        agent_ids: List[str],
+        collaboration_pattern: str = "sequential"
+    ) -> Dict[str, Any]:
+        """Setup collaboration between Pydantic AI agents"""
+        
+        agents = {aid: self.agents[aid] for aid in agent_ids if aid in self.agents}
+        
+        if len(agents) < 2:
+            return {
+                "status": "error",
+                "message": "Need at least 2 agents for collaboration"
+            }
+        
+        # Implement different collaboration patterns
+        if collaboration_pattern == "sequential":
+            # Agents work in sequence
+            async def sequential_collaboration(task: str) -> Dict[str, Any]:
+                results = []
+                current_input = task
+                
+                for agent_id, agent in agents.items():
+                    result = await agent.run_agent(current_input)
+                    results.append({
+                        "agent_id": agent_id,
+                        "result": result
+                    })
+                    
+                    # Use result as input for next agent
+                    if result.get("status") == "success":
+                        current_input = result.get("result", "")
+                
+                return {
+                    "status": "success",
+                    "collaboration_type": "sequential",
+                    "results": results,
+                    "final_result": current_input
+                }
+        
+        elif collaboration_pattern == "parallel":
+            # Agents work in parallel
+            async def parallel_collaboration(task: str) -> Dict[str, Any]:
+                tasks = [
+                    agent.run_agent(task) 
+                    for agent in agents.values()
+                ]
+                
+                results = await asyncio.gather(*tasks, return_exceptions=True)
+                
+                return {
+                    "status": "success",
+                    "collaboration_type": "parallel",
+                    "results": [
+                        {
+                            "agent_id": agent_id,
+                            "result": result
+                        }
+                        for agent_id, result in zip(agent_ids, results)
+                    ]
+                }
+        
+        else:
+            return {
+                "status": "error",
+                "message": f"Unknown collaboration pattern: {collaboration_pattern}"
+            }
+        
+        # Return the collaboration function
+        return {
+            "status": "success",
+            "collaboration_function": locals().get(f"{collaboration_pattern}_collaboration"),
+            "agents": list(agents.keys()),
+            "pattern": collaboration_pattern
+        }
+
+# Export classes for easy importing
+__all__ = [
+    'PydanticAIAgentConfig',
+    'PydanticAIMCPAgent',
+    'PydanticAIAgentBridge'
+]