agent-mcp 0.1.3__py3-none-any.whl → 0.1.4__py3-none-any.whl

agent_mcp/__init__.py

@@ -2,6 +2,8 @@
 AgentMCP - Model Context Protocol for AI Agents
 """
 
+__version__ = "0.1.4"
+
 from .mcp_agent import MCPAgent
 from .mcp_decorator import mcp_agent
 from .enhanced_mcp_agent import EnhancedMCPAgent
@@ -12,5 +14,3 @@ from .heterogeneous_group_chat import HeterogeneousGroupChat
 from .langchain_mcp_adapter import LangchainMCPAdapter
 from .crewai_mcp_adapter import CrewAIMCPAdapter
 from .langgraph_mcp_adapter import LangGraphMCPAdapter
-
-__version__ = "0.1.2"

agent_mcp/camel_mcp_adapter.py

@@ -0,0 +1,521 @@
+"""
+Adapter for Camel AI agents to work with MCP.
+"""
+
+import asyncio
+from typing import Dict, Any, Optional
+from .mcp_agent import MCPAgent
+from .mcp_transport import MCPTransport
+from camel.agents import ChatAgent  # Import the core Camel AI agent
+import traceback
+import json
+import uuid
+
+# --- Setup Logger ---
+import logging
+logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
+logger = logging.getLogger(__name__)
+# --- End Logger Setup ---
+
+class CamelMCPAdapter(MCPAgent):
+    """Adapter for Camel AI ChatAgent to work with MCP"""
+
+    def __init__(self,
+                 name: str,
+                 transport: Optional[MCPTransport] = None,
+                 client_mode: bool = False,
+                 camel_agent: ChatAgent = None,  # Expect a Camel ChatAgent instance
+                 system_message: str = "",
+                 **kwargs):
+        # Set default system message if none provided and agent doesn't have one
+        if not system_message and camel_agent and hasattr(camel_agent, 'system_message') and not camel_agent.system_message:
+            system_message = "I am a Camel AI agent ready to assist."
+        elif not system_message:
+             system_message = "I am a Camel AI agent ready to assist."
+
+        # Initialize parent with system message
+        # Use provided system_message or agent's if available
+        effective_system_message = system_message or (camel_agent.system_message.content if camel_agent and camel_agent.system_message else "Camel AI Assistant")
+        super().__init__(name=name, system_message=effective_system_message, **kwargs)
+
+        # Set instance attributes
+        self.transport = transport
+        self.client_mode = client_mode
+        self.camel_agent = camel_agent # Store the Camel ChatAgent
+        self.task_queue = asyncio.Queue()
+        self._task_processor = None
+        self._message_processor = None
+        self._processed_tasks = set()  # For idempotency check
+
+        if not self.camel_agent:
+            raise ValueError("A camel.agents.ChatAgent instance must be provided.")
+
+    async def connect_to_server(self, server_url: str):
+        """Connect to another agent's server"""
+        if not self.client_mode or not self.transport:
+            raise ValueError("Agent not configured for client mode")
+
+        # Register with the server
+        registration = {
+            "type": "registration",
+            "agent_id": self.mcp_id,
+            "name": self.name,
+            "capabilities": [] # Define capabilities based on agent if needed
+        }
+
+        response = await self.transport.send_message(server_url, registration)
+        if response.get("status") == "ok":
+            print(f"Successfully connected to server at {server_url}")
+# transport message format
+    async def handle_incoming_message(self, message: Dict[str, Any], message_id: Optional[str] = None):
+        """Handle incoming messages from other agents"""
+        # First check if type is directly in the message
+        msg_type = message.get("type")
+        logger.info(f"[{self.name}] Raw message: {message}")
+
+        # If not, check if it's inside the content field
+        if not msg_type and "content" in message and isinstance(message["content"], dict):
+            msg_type = message["content"].get("type")
+
+        # Extract sender with nested JSON support
+        sender = self._extract_sender(message)
+        task_id = message.get("task_id") or message.get("content", {}).get("task_id") if isinstance(message.get("content"), dict) else message.get("task_id")
+        logger.info(f"[{self.name}] Received message (ID: {message_id}) of type '{msg_type}' from {sender} (Task ID: {task_id})")
+
+        # --- Idempotency Check ---
+        if not super()._should_process_message(message):
+            # If skipped, acknowledge and stop
+            if message_id and self.transport:
+                asyncio.create_task(self.transport.acknowledge_message(self.name, message_id))
+                logger.info(f"[{self.name}] Acknowledged duplicate task {task_id} (msg_id: {message_id})")
+            return
+        # --- End Idempotency Check ---
+
+        if msg_type == "task":
+            logger.info(f"[{self.name}] Queueing task {task_id} (message_id: {message_id}) from {sender}")
+            content = message.get("content", {})
+            # Ensure task_id and description are retrieved correctly from nested content if necessary
+            current_task_id = content.get("task_id", message.get("task_id"))
+            description = content.get("description", message.get("description"))
+            reply_to = content.get("reply_to", message.get("reply_to"))
+
+            if not current_task_id or not description:
+                logger.error(f"[{self.name}] Task message missing required fields (task_id or description): {message}")
+                # Acknowledge if possible, even if invalid, to prevent reprocessing
+                if message_id and self.transport:
+                     asyncio.create_task(self.transport.acknowledge_message(self.name, message_id))
+                return
+
+            # Add message_id to task context for later acknowledgement
+            task_context = {
+                "type": "task", # Explicitly set type for process_tasks
+                "task_id": current_task_id,
+                "description": description,
+                "reply_to": reply_to,
+                "sender": sender,
+                "message_id": message_id, # Store message_id for acknowledgement
+                "original_message": message # Store original for context if needed
+            }
+
+            logger.debug(f"[{self.name}] Queueing task context: {task_context}")
+            await self.task_queue.put(task_context)
+            logger.debug(f"[{self.name}] Successfully queued task {current_task_id}")
+
+        elif msg_type == "task_result":
+            # Received a result, treat it as the next step in the conversation
+            result_content = message.get("result")
+            
+            # --- Robust extraction for various formats ---
+            content = message.get("content")
+            if result_content is None and content is not None:
+                # 1. Try content["result"]
+                if isinstance(content, dict) and "result" in content:
+                    result_content = content["result"]
+                # 2. Try content["text"] as JSON
+                elif isinstance(content, dict) and "text" in content:
+                    text_val = content["text"]
+                    if isinstance(text_val, str):
+                        try:
+                            parsed = json.loads(text_val)
+                            if isinstance(parsed, dict) and "result" in parsed:
+                                result_content = parsed["result"]
+                        except Exception:
+                            pass
+                # 3. Try content itself as JSON string
+                elif isinstance(content, str):
+                    try:
+                        parsed = json.loads(content)
+                        if isinstance(parsed, dict) and "result" in parsed:
+                            result_content = parsed["result"]
+                    except Exception:
+                        pass
+                # 4. Fallback: use content["text"] as plain string
+                if result_content is None and isinstance(content, dict) and "text" in content:
+                    result_content = content["text"]
+
+            # Add direct parsing of content["text"] structure
+            if isinstance(result_content, str):
+                try:
+                    text_content = json.loads(result_content)
+                    if isinstance(text_content, dict):
+                        result_content = text_content
+                except json.JSONDecodeError:
+                    pass
+
+            # Handle JSON string content
+            if isinstance(result_content, str):
+                try:
+                    result_content = json.loads(result_content)
+                except json.JSONDecodeError:
+                    pass
+
+            original_task_id = (
+                (result_content.get("task_id") if isinstance(result_content, dict) else None)
+                or message.get("task_id")
+            )
+            logger.info(f"[{self.name}] Received task_result from {sender} for task {original_task_id}. Content: '{str(result_content)[:100]}...'")
+
+            if not result_content:
+                logger.warning(f"[{self.name}] Received task_result from {sender} with empty content.")
+                # Acknowledge the result message even if content is empty
+                if message_id and self.transport:
+                    asyncio.create_task(self.transport.acknowledge_message(self.name, message_id))
+                return
+
+            # Create a *new* task for this agent based on the received result
+            new_task_id = f"conv_{uuid.uuid4()}" # Generate a new ID for this conversational turn
+            new_task_context = {
+                "type": "task", # Still a task for this agent to process
+                "task_id": new_task_id,
+                "description": str(result_content), # The result becomes the new input/description
+                "reply_to": message.get("reply_to") or result_content.get("reply_to"),
+                "sender": sender, # This agent is the conceptual sender of this internal task
+                "message_id": message_id # Carry over original message ID for acknowledgement
+            }
+
+            logger.info(f"[{self.name}] Queueing new conversational task {new_task_id} based on result from {sender}")
+            await self.task_queue.put(new_task_context)
+            logger.debug(f"[{self.name}] Successfully queued new task {new_task_id}")
+
+        else:
+            logger.warning(f"[{self.name}] Received unknown message type: {msg_type}. Message: {message}")
+            # Acknowledge non-task messages immediately if they have an ID
+            if message_id and self.transport:
+                 asyncio.create_task(self.transport.acknowledge_message(self.name, message_id))
+
+# message processor loop making sure the message is a dictionary and then processing it
+    async def process_messages(self):
+        logger.info(f"[{self.name}] Message processor loop started.")
+        while True:
+            try:
+                logger.debug(f"[{self.name}] Waiting for message from transport...")
+                # Get message from transport (without agent_name parameter)
+                message, message_id = await self.transport.receive_message()
+                logger.debug(f"[{self.name}] Received raw message from transport: {message} (ID: {message_id})")
+
+
+                if message is None:
+                    logger.debug(f"[{self.name}] Received None message, continuing loop...")
+                    await asyncio.sleep(0.1) # Avoid busy-waiting
+                    continue
+
+                # Ensure message is a dictionary before proceeding
+                if not isinstance(message, dict):
+                    logger.error(f"[{self.name}] Received non-dict message: {message} (type: {type(message)}), skipping.")
+                    # Attempt to acknowledge if possible
+                    if message_id and self.transport:
+                         asyncio.create_task(self.transport.acknowledge_message(self.name, message_id))
+                    continue
+
+                logger.info(f"[{self.name}] Processing message {message_id}: {message}")
+                await self.handle_incoming_message(message, message_id)
+
+            except asyncio.CancelledError:
+                logger.info(f"[{self.name}] Message processor cancelled.")
+                break
+            except Exception as e:
+                logger.error(f"[{self.name}] Error in message processor: {e}", exc_info=True)
+                traceback.print_exc()
+                # Avoid immediate exit on error, maybe add a delay
+                await asyncio.sleep(1)
+        logger.info(f"[{self.name}] Message processor loop finished.")
+
+
+    async def process_tasks(self):
+        logger.info(f"[{self.name}] Task processor loop started.")
+        while True:
+            try:
+                logger.debug(f"[{self.name}] Waiting for task from queue...")
+                task_context = await self.task_queue.get()
+                logger.info(f"\n[{self.name}] Dequeued item from task queue: {task_context}")
+
+                if not isinstance(task_context, dict):
+                    logger.error(f"[{self.name}] Task item is not a dictionary: {task_context}")
+                    self.task_queue.task_done()
+                    continue
+
+                # Extract task details from the context dictionary
+                task_desc = task_context.get("description")
+                task_id = task_context.get("task_id")
+                reply_to = task_context.get("reply_to")
+                message_id = task_context.get("message_id") # Retrieve message_id for ack
+                task_type = task_context.get("type") # Should be 'task'
+
+                logger.info(f"[{self.name}] Processing task details:")
+                logger.info(f"  - Task ID: {task_id}")
+                logger.info(f"  - Type: {task_type}")
+                logger.info(f"  - Reply To: {reply_to}")
+                logger.info(f"  - Description: {str(task_desc)[:100]}...")
+                logger.info(f"  - Original Message ID: {message_id}")
+
+                if not task_desc or not task_id:
+                    logger.error(f"[{self.name}] Task context missing description or task_id: {task_context}")
+                    self.task_queue.task_done()
+                     # Acknowledge message even if task context is bad
+                    if message_id and self.transport:
+                         asyncio.create_task(self.transport.acknowledge_message(self.name, message_id))
+                         logger.warning(f"[{self.name}] Acknowledged message {message_id} for invalid task context.")
+                    continue
+                
+                # Ensure it's actually a task we should process
+                if task_type != "task":
+                    logger.error(f"[{self.name}] Invalid item type in task queue: {task_type}. Item: {task_context}")
+                    self.task_queue.task_done()
+                    if message_id and self.transport:
+                        asyncio.create_task(self.transport.acknowledge_message(self.name, message_id))
+                    continue
+
+                logger.info(f"[{self.name}] Starting execution of task {task_id}: '{str(task_desc)[:50]}...'" )
+
+                # --- Execute task using Camel AI agent ---
+                result_str = None
+                try:
+                    logger.debug(f"[{self.name}] Calling camel_agent.astep with task description...")
+                    # Use astep for asynchronous execution if available, otherwise step
+                    if hasattr(self.camel_agent, 'astep'):
+                        response = await self.camel_agent.astep(task_desc)
+                    else:
+                        # Fallback to synchronous step in executor if astep not available
+                        # Note: This blocks the task processor loop. Consider running in executor.
+                        logger.warning(f"[{self.name}] Camel agent does not have 'astep'. Using synchronous 'step'. This may block.")
+                        # For safety, run sync 'step' in an executor to avoid blocking asyncio loop
+                        loop = asyncio.get_running_loop()
+                        response = await loop.run_in_executor(None, self.camel_agent.step, task_desc)
+
+
+                    # Extract content from the response object
+                    # Assuming response has a 'msg' attribute with 'content' based on docs example
+                    if response and hasattr(response, 'msg') and hasattr(response.msg, 'content'):
+                        result_str = response.msg.content
+                        logger.debug(f"[{self.name}] Camel agent execution successful. Result: '{str(result_str)[:100]}...'" )
+                    elif response and hasattr(response, 'messages') and response.messages:
+                        # Handle cases where response might be a list of messages (e.g., RolePlaying)
+                        # Take the last message content as the result for simplicity
+                        last_msg = response.messages[-1]
+                        if hasattr(last_msg, 'content'):
+                           result_str = last_msg.content
+                           logger.debug(f"[{self.name}] Camel agent execution successful (from messages). Result: '{str(result_str)[:100]}...'" )
+                        else:
+                            logger.warning(f"[{self.name}] Camel agent response's last message has no 'content'. Response: {response}")
+                            result_str = str(response) # Fallback
+                    elif response:
+                         logger.warning(f"[{self.name}] Camel agent response format unexpected. Using str(response). Response: {response}")
+                         result_str = str(response) # Fallback to string representation
+                    else:
+                        logger.warning(f"[{self.name}] Camel agent returned None or empty response.")
+                        result_str = "Agent returned no response."
+
+
+                except Exception as e:
+                    logger.error(f"[{self.name}] Camel agent execution failed for task {task_id}: {e}", exc_info=True)
+                    traceback.print_exc()
+                    result_str = f"Agent execution failed due to an error: {str(e)}"
+
+                # Ensure result is always a string before sending
+                if not isinstance(result_str, str):
+                    try:
+                        result_str = json.dumps(result_str) # Try serializing complex types
+                    except (TypeError, OverflowError):
+                        result_str = str(result_str) # Fallback
+
+                logger.info(f"[{self.name}] Sending result for task {task_id} to {reply_to}")
+                # --- Send the result back ---
+                if reply_to and self.transport:
+                    try:
+                        # Extract target agent name from reply_to (assuming format http://.../agent_name)
+                        target_agent_name = reply_to # Default if parsing fails
+                        try:
+                            target_agent_name = reply_to.split('/')[-1]
+                            if not target_agent_name: # Handle trailing slash case
+                                target_agent_name = reply_to.split('/')[-2]
+                        except IndexError:
+                            logger.warning(f"[{self.name}] Could not reliably extract agent name from reply_to URL: {reply_to}. Using full URL.")
+
+
+                        logger.debug(f"[{self.name}] Sending result to target agent: {target_agent_name} (extracted from {reply_to})")
+
+                        await self.transport.send_message(
+                            target_agent_name, # Send to the specific agent name/ID
+                            {
+                                "type": "task_result",
+                                "task_id": task_id,
+                                "result": result_str,
+                                "sender": self.name,
+                                "reply_to": self.name,  # Add this so the receiving agent knows where to send follow-up messages
+                                "original_message_id": message_id # Include original message ID for tracing/ack
+                            }
+                        )
+                        logger.debug(f"[{self.name}] Result sent successfully for task {task_id}")
+
+                        # Acknowledge task completion *after* sending result, using message_id
+                        if message_id:
+                            await self.transport.acknowledge_message(self.name, message_id)
+                            logger.info(f"[{self.name}] Task {task_id} acknowledged via message_id {message_id}")
+                        else:
+                            logger.warning(f"[{self.name}] No message_id found for task {task_id} in context, cannot acknowledge transport message.")
+
+                    except Exception as send_error:
+                        logger.error(f"[{self.name}] Failed to send result or acknowledge for task {task_id}: {send_error}", exc_info=True)
+                        traceback.print_exc()
+                        # Decide if task should be retried or marked done despite send failure
+                        # For now, mark done to avoid loop, but log error clearly
+                else:
+                    logger.warning(f"[{self.name}] No reply_to address or transport configured for task {task_id}. Cannot send result or acknowledge.")
+                    # If no reply_to, we might still need to acknowledge if we have a message_id
+                    if message_id and self.transport:
+                         logger.warning(f"[{self.name}] Acknowledging message {message_id} for task {task_id} even though result wasn't sent (no reply_to)." )
+                         await self.transport.acknowledge_message(self.name, message_id)
+
+
+                # Mark task as completed internally *after* processing and attempting send/ack
+                super()._mark_task_completed(task_id) # Call base class method
+
+                self.task_queue.task_done()
+                logger.info(f"[{self.name}] Task {task_id} fully processed and marked done.")
+
+            except asyncio.CancelledError:
+                logger.info(f"[{self.name}] Task processor cancelled.")
+                break
+            except Exception as e:
+                logger.error(f"[{self.name}] Unhandled error in task processing loop: {e}", exc_info=True)
+                traceback.print_exc()
+                # Ensure task_done is called even in unexpected error to prevent queue blockage
+                try:
+                    self.task_queue.task_done()
+                except ValueError: # Already done
+                    pass
+                await asyncio.sleep(1) # Prevent fast error loop
+
+        logger.info(f"[{self.name}] Task processor loop finished.")
+
+    # --- Helper methods inherited or overridden from MCPAgent ---
+    # _should_process_message and _mark_task_completed are inherited from MCPAgent
+    # but shown here for clarity from the original langchain adapter template.
+    # We rely on the super() calls within handle_incoming_message and process_tasks.
+
+    # def _should_process_message(self, message: Dict[str, Any]) -> bool:
+    #     """Check if a message should be processed based on idempotency"""
+    #     # Implementation relies on MCPAgent's base method via super()
+
+    # def _mark_task_completed(self, task_id: str) -> None:
+    #     """Mark a task as completed for idempotency tracking"""
+    #     # Implementation relies on MCPAgent's base method via super()
+
+    async def run(self):
+        """Start the agent's message and task processing loops."""
+        if not self.transport:
+            raise ValueError("Transport must be configured to run the agent.")
+
+        logger.info(f"[{self.name}] Starting CamelMCPAdapter run loop...")
+
+        # Start the message processing loop
+        self._message_processor = asyncio.create_task(self.process_messages())
+        # Start the task processing loop
+        self._task_processor = asyncio.create_task(self.process_tasks())
+
+        try:
+            # Keep running until tasks are cancelled
+            await asyncio.gather(self._message_processor, self._task_processor)
+        except asyncio.CancelledError:
+            logger.info(f"[{self.name}] Agent run loop cancelled.")
+        finally:
+            logger.info(f"[{self.name}] Agent run loop finished.")
+
+    async def stop(self):
+        """Stop the agent's processing loops gracefully."""
+        logger.info(f"[{self.name}] Stopping agent...")
+        if self._message_processor and not self._message_processor.done():
+            self._message_processor.cancel()
+        if self._task_processor and not self._task_processor.done():
+            self._task_processor.cancel()
+
+        # Wait for tasks to finish cancelling
+        try:
+            await asyncio.gather(self._message_processor, self._task_processor, return_exceptions=True)
+        except asyncio.CancelledError:
+            pass # Expected
+        logger.info(f"[{self.name}] Agent stopped.")
+
+# Example Usage (for demonstration purposes, would normally be in a separate script)
+# async def main():
+#     # Requires setting up a transport (e.g., MCPTransport)
+#     # Requires creating a Camel ChatAgent instance
+#     from camel.models import ModelFactory
+#     from camel.types import ModelType
+#     import os
+#
+#     # Ensure API key is set
+#     # os.environ["OPENAI_API_KEY"] = "your_key_here"
+#
+#     # 1. Create a Camel ChatAgent
+#     model = ModelFactory.create(model_type=ModelType.GPT_4O_MINI)
+#     camel_agent_instance = ChatAgent(system_message="You are a helpful geography expert.", model=model)
+#
+#     # 2. Create a transport (replace with actual transport implementation)
+#     class MockTransport(MCPTransport): # Replace with your actual transport
+#         async def send_message(self, target: str, message: Dict[str, Any]):
+#             print(f"[MockTransport] Sending to {target}: {message}")
+#             # Simulate response for registration
+#             if message.get("type") == "registration":
+#                 return {"status": "ok"}
+#             return {"status": "sent"}
+#         async def receive_message(self):
+#             print("[MockTransport] receive_message called, simulating task...")
+#             await asyncio.sleep(5) # Simulate delay
+#             # Simulate receiving a task message
+#             return {
+#                 "type": "task",
+#                 "task_id": "task-123",
+#                 "description": "What is the capital of France?",
+#                 "sender": "user_agent",
+#                 "reply_to": "http://localhost:8000/user_agent" # Example reply URL
+#             }, "msg-abc" # Message and ID
+#         async def acknowledge_message(self, agent_id: str, message_id: str):
+#              print(f"[MockTransport] Acknowledging message {message_id} for agent {agent_id}")
+#         async def connect(self): pass
+#         async def disconnect(self): pass
+#
+#     transport_instance = MockTransport()
+#
+#     # 3. Create the adapter
+#     adapter = CamelMCPAdapter(
+#         name="CamelGeographyAgent",
+#         transport=transport_instance,
+#         camel_agent=camel_agent_instance
+#     )
+#
+#     # 4. Run the adapter
+#     print("Starting CamelMCPAdapter...")
+#     try:
+#         await adapter.run()
+#     except KeyboardInterrupt:
+#         await adapter.stop()
+#     print("Adapter finished.")
+#
+# if __name__ == "__main__":
+#      # Note: Requires OPENAI_API_KEY environment variable to be set
+#      # if os.getenv("OPENAI_API_KEY"):
+#      #    asyncio.run(main())
+#      # else:
+#      #    print("Please set the OPENAI_API_KEY environment variable to run the example.")
+#      pass # Keep example code commented out in the main file 

agent_mcp/cli.py

@@ -0,0 +1,47 @@
+"""
+Command-line interface for the Agent MCP package.
+"""
+import argparse
+import logging
+from typing import Optional
+
+def main():
+    """Entry point for the command-line interface."""
+    parser = argparse.ArgumentParser(description="Agent MCP - Multi-agent Collaboration Platform")
+    parser.add_argument(
+        "--version", 
+        action="store_true", 
+        help="Show version information"
+    )
+    parser.add_argument(
+        "-v", 
+        "--verbose", 
+        action="count", 
+        default=0, 
+        help="Increase verbosity (use -vv for debug level)"
+    )
+    
+    args = parser.parse_args()
+    
+    # Configure logging
+    log_level = logging.WARNING
+    if args.verbose == 1:
+        log_level = logging.INFO
+    elif args.verbose >= 2:
+        log_level = logging.DEBUG
+    
+    logging.basicConfig(
+        level=log_level,
+        format="%(asctime)s - %(name)s - %(levelname)s - %(message)s"
+    )
+    
+    if args.version:
+        from agent_mcp import __version__
+        print(f"Agent MCP version {__version__}")
+        return
+    
+    # Default action (you can add more commands here)
+    print("Agent MCP - Use --help for usage information")
+
+if __name__ == "__main__":
+    main()