mbxai 0.6.10__tar.gz → 0.6.11__tar.gz

{mbxai-0.6.10 → mbxai-0.6.11}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mbxai
-Version: 0.6.10
+Version: 0.6.11
 Summary: MBX AI SDK
 Project-URL: Homepage, https://www.mibexx.de
 Project-URL: Documentation, https://www.mibexx.de

{mbxai-0.6.10 → mbxai-0.6.11}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "mbxai"
-version = "0.6.10"
+version = "0.6.11"
 authors = [
     { name = "MBX AI" }
 ]

{mbxai-0.6.10 → mbxai-0.6.11}/setup.py

@@ -2,7 +2,7 @@ from setuptools import setup, find_packages
 
 setup(
     name="mbxai",
-    version="0.6.10",
+    version="0.6.11",
     author="MBX AI",
     description="MBX AI SDK",
     long_description=open("README.md").read(),

{mbxai-0.6.10 → mbxai-0.6.11}/src/mbxai/__init__.py

@@ -2,4 +2,4 @@
 MBX AI package.
 """
 
-__version__ = "0.6.0" 
+__version__ = "0.5.25" 

{mbxai-0.6.10 → mbxai-0.6.11}/src/mbxai/mcp/server.py

@@ -31,7 +31,7 @@ class MCPServer:
         self.app = FastAPI(
             title=self.name,
             description=self.description,
-            version="0.6.10",
+            version="0.5.25",
         )
         
         # Initialize MCP server

mbxai-0.6.11/src/mbxai/tools/client.py

@@ -0,0 +1,425 @@
+"""
+ToolClient implementation for MBX AI.
+"""
+
+from typing import Any, Callable, TypeVar, cast
+import logging
+import inspect
+import json
+from pydantic import BaseModel
+from ..openrouter import OpenRouterClient
+from .types import Tool, ToolCall
+import asyncio
+
+logger = logging.getLogger(__name__)
+
+T = TypeVar("T", bound=BaseModel)
+
+class ToolClient:
+    """Client for handling tool calls with OpenRouter."""
+
+    def __init__(self, openrouter_client: OpenRouterClient) -> None:
+        """Initialize the ToolClient.
+
+        Args:
+            openrouter_client: The OpenRouter client to use
+        """
+        self._client = openrouter_client
+        self._tools: dict[str, Tool] = {}
+
+    def register_tool(
+        self,
+        name: str,
+        description: str,
+        function: Callable[..., Any],
+        schema: dict[str, Any],
+    ) -> None:
+        """Register a new tool.
+
+        Args:
+            name: The name of the tool
+            description: A description of what the tool does
+            function: The function to call when the tool is used
+            schema: The JSON schema for the tool's parameters
+        """
+        tool = Tool(
+            name=name,
+            description=description,
+            function=function,
+            schema=schema,
+        )
+        self._tools[name] = tool
+        logger.info(f"Registered tool: {name}")
+
+    def _truncate_content(self, content: str | None, max_length: int = 100) -> str:
+        """Truncate content for logging."""
+        if not content:
+            return "None"
+        if len(content) <= max_length:
+            return content
+        return content[:max_length] + "..."
+
+    def _truncate_dict(self, data: dict[str, Any], max_length: int = 50) -> str:
+        """Truncate dictionary values for logging."""
+        if not data:
+            return "{}"
+        truncated = {}
+        for k, v in data.items():
+            if isinstance(v, str):
+                truncated[k] = self._truncate_content(v, max_length)
+            elif isinstance(v, dict):
+                truncated[k] = self._truncate_dict(v, max_length)
+            else:
+                truncated[k] = str(v)[:max_length] + "..." if len(str(v)) > max_length else v
+        return str(truncated)
+
+    def _validate_message_sequence(self, messages: list[dict[str, Any]], validate_responses: bool = True) -> None:
+        """Validate the message sequence for tool calls and responses.
+        
+        Args:
+            messages: The message sequence to validate
+            validate_responses: Whether to validate that all tool calls have responses
+        """
+        tool_call_ids = set()
+        tool_response_ids = set()
+        
+        for i, msg in enumerate(messages):
+            role = msg.get("role")
+            if role == "assistant" and "tool_calls" in msg:
+                # Track tool calls
+                for tc in msg["tool_calls"]:
+                    tool_call_ids.add(tc["id"])
+                    logger.info(f"Found tool call {tc['id']} for {tc['function']['name']} in message {i}")
+            elif role == "tool":
+                # Track tool responses
+                tool_response_ids.add(msg["tool_call_id"])
+                logger.info(f"Found tool response for call ID {msg['tool_call_id']} in message {i}")
+        
+        # Only validate responses if requested
+        if validate_responses:
+            # Check for missing responses
+            missing_responses = tool_call_ids - tool_response_ids
+            if missing_responses:
+                logger.error(f"Missing tool responses for call IDs: {missing_responses}")
+                logger.error("Message sequence:")
+                for i, msg in enumerate(messages):
+                    role = msg.get("role", "unknown")
+                    if role == "assistant" and "tool_calls" in msg:
+                        logger.error(f"  Message {i} - Assistant with tool calls: {[tc['id'] for tc in msg['tool_calls']]}")
+                    elif role == "tool":
+                        logger.error(f"  Message {i} - Tool response for call ID: {msg['tool_call_id']}")
+                    else:
+                        logger.error(f"  Message {i} - {role}: {self._truncate_content(msg.get('content'))}")
+                raise ValueError(f"Invalid message sequence: missing responses for tool calls {missing_responses}")
+
+    def _log_messages(self, messages: list[dict[str, Any]], validate_responses: bool = True) -> None:
+        """Log the messages being sent to OpenRouter.
+        
+        Args:
+            messages: The messages to log
+            validate_responses: Whether to validate that all tool calls have responses
+        """
+        logger.info("Sending messages to OpenRouter:")
+        for i, msg in enumerate(messages):
+            role = msg.get("role", "unknown")
+            content = self._truncate_content(msg.get("content"))
+            tool_calls = msg.get("tool_calls", [])
+            tool_call_id = msg.get("tool_call_id")
+            
+            if tool_calls:
+                tool_call_info = [
+                    f"{tc['function']['name']}(id={tc['id']})"
+                    for tc in tool_calls
+                ]
+                logger.info(f"  Message {i} - {role}: content='{content}', tool_calls={tool_call_info}")
+            elif tool_call_id:
+                logger.info(f"  Message {i} - {role}: content='{content}', tool_call_id={tool_call_id}")
+            else:
+                logger.info(f"  Message {i} - {role}: content='{content}'")
+        
+        # Validate message sequence
+        self._validate_message_sequence(messages, validate_responses)
+
+    async def _process_tool_calls(self, message: Any, messages: list[dict[str, Any]]) -> None:
+        """Process all tool calls in a message.
+        
+        Args:
+            message: The message containing tool calls
+            messages: The list of messages to add responses to
+        """
+        if not message.tool_calls:
+            return
+
+        # Process all tool calls first
+        tool_responses = []
+        for tool_call in message.tool_calls:
+            tool = self._tools.get(tool_call.function.name)
+            if not tool:
+                raise ValueError(f"Unknown tool: {tool_call.function.name}")
+
+            # Parse arguments if they're a string
+            arguments = tool_call.function.arguments
+            if isinstance(arguments, str):
+                try:
+                    arguments = json.loads(arguments)
+                except json.JSONDecodeError as e:
+                    logger.error(f"Failed to parse tool arguments: {e}")
+                    raise ValueError(f"Invalid tool arguments format: {arguments}")
+
+            # Call the tool
+            logger.info(f"Calling tool: {tool.name} with args: {self._truncate_dict(arguments)}")
+            if inspect.iscoroutinefunction(tool.function):
+                result = await tool.function(**arguments)
+            else:
+                result = tool.function(**arguments)
+
+            # Convert result to JSON string if it's not already
+            if not isinstance(result, str):
+                result = json.dumps(result)
+
+            # Create the tool response
+            tool_response = {
+                "role": "tool",
+                "tool_call_id": tool_call.id,
+                "content": result,
+            }
+            tool_responses.append(tool_response)
+            logger.info(f"Created tool response for call ID {tool_call.id}")
+
+        # Add all tool responses to the messages
+        messages.extend(tool_responses)
+        logger.info(f"Message count: {len(messages)}, Added {len(tool_responses)} tool responses to messages")
+
+        # Validate the message sequence
+        self._validate_message_sequence(messages, validate_responses=True)
+
+        # Log the messages we're about to send
+        self._log_messages(messages, validate_responses=False)
+
+    async def chat(
+        self,
+        messages: list[dict[str, Any]],
+        *,
+        model: str | None = None,
+        stream: bool = False,
+        **kwargs: Any,
+    ) -> Any:
+        """Chat with the model, handling tool calls."""
+        tools = [tool.to_openai_function() for tool in self._tools.values()]
+        
+        if tools:
+            logger.info(f"Available tools: {[tool['function']['name'] for tool in tools]}")
+            kwargs["tools"] = tools
+            kwargs["tool_choice"] = "auto"
+        
+        while True:
+            # Get the model's response
+            response = self._client.chat_completion(
+                messages=messages,
+                model=model,
+                stream=stream,
+                **kwargs,
+            )
+
+            if stream:
+                return response
+
+            message = response.choices[0].message
+            # Add the assistant's message with tool calls
+            assistant_message = {
+                "role": "assistant",
+                "content": message.content or None,  # Ensure content is None if empty
+            }
+            if message.tool_calls:
+                assistant_message["tool_calls"] = [
+                    {
+                        "id": tool_call.id,
+                        "type": "function",
+                        "function": {
+                            "name": tool_call.function.name,
+                            "arguments": tool_call.function.arguments,
+                        },
+                    }
+                    for tool_call in message.tool_calls
+                ]
+            messages.append(assistant_message)
+            logger.info(f"Message count: {len(messages)}, Added assistant message with tool calls: {[tc.function.name for tc in message.tool_calls] if message.tool_calls else None}")
+
+            # If there are no tool calls, we're done
+            if not message.tool_calls:
+                return response
+
+            # Process all tool calls
+            tool_responses = []
+            for tool_call in message.tool_calls:
+                tool = self._tools.get(tool_call.function.name)
+                if not tool:
+                    raise ValueError(f"Unknown tool: {tool_call.function.name}")
+
+                # Parse arguments if they're a string
+                arguments = tool_call.function.arguments
+                if isinstance(arguments, str):
+                    try:
+                        arguments = json.loads(arguments)
+                    except json.JSONDecodeError as e:
+                        logger.error(f"Failed to parse tool arguments: {e}")
+                        raise ValueError(f"Invalid tool arguments format: {arguments}")
+
+                # Call the tool
+                logger.info(f"Calling tool: {tool.name} with args: {self._truncate_dict(arguments)}")
+                try:
+                    if inspect.iscoroutinefunction(tool.function):
+                        result = await asyncio.wait_for(tool.function(**arguments), timeout=300.0)  # 5 minutes timeout
+                    else:
+                        result = tool.function(**arguments)
+                    logger.info(f"Tool {tool.name} completed successfully")
+                except asyncio.TimeoutError:
+                    logger.error(f"Tool {tool.name} timed out after 5 minutes")
+                    result = {"error": "Tool execution timed out after 5 minutes"}
+                except Exception as e:
+                    logger.error(f"Error calling tool {tool.name}: {str(e)}")
+                    result = {"error": f"Tool execution failed: {str(e)}"}
+
+                # Convert result to JSON string if it's not already
+                if not isinstance(result, str):
+                    result = json.dumps(result)
+
+                # Create the tool response
+                tool_response = {
+                    "role": "tool",
+                    "tool_call_id": tool_call.id,
+                    "content": result,
+                }
+                tool_responses.append(tool_response)
+                logger.info(f"Created tool response for call ID {tool_call.id}")
+
+            # Add all tool responses to the messages
+            messages.extend(tool_responses)
+            logger.info(f"Message count: {len(messages)}, Added {len(tool_responses)} tool responses to messages")
+
+            # Validate the message sequence
+            self._validate_message_sequence(messages, validate_responses=True)
+
+            # Log the messages we're about to send
+            self._log_messages(messages, validate_responses=False)
+
+            # Continue the loop to get the next response
+            continue
+
+    async def parse(
+        self,
+        messages: list[dict[str, Any]],
+        response_format: type[T],
+        *,
+        model: str | None = None,
+        stream: bool = False,
+        **kwargs: Any,
+    ) -> Any:
+        """Chat with the model and parse the response into a Pydantic model.
+
+        Args:
+            messages: The conversation messages
+            response_format: The Pydantic model to parse the response into
+            model: Optional model override
+            stream: Whether to stream the response
+            **kwargs: Additional parameters for the chat completion
+
+        Returns:
+            The parsed response from the model
+        """
+        tools = [tool.to_openai_function() for tool in self._tools.values()]
+        
+        if tools:
+            logger.info(f"Available tools: {[tool['function']['name'] for tool in tools]}")
+            kwargs["tools"] = tools
+            kwargs["tool_choice"] = "auto"
+        
+        while True:
+            # Log messages before sending to OpenRouter
+            self._log_messages(messages)
+            
+            # Get the model's response
+            response = self._client.chat_completion_parse(
+                messages=messages,
+                response_format=response_format,
+                model=model,
+                stream=stream,
+                **kwargs,
+            )
+
+            if stream:
+                return response
+
+            message = response.choices[0].message
+            # Add the assistant's message with tool calls
+            assistant_message = {
+                "role": "assistant",
+                "content": message.content or None,  # Ensure content is None if empty
+            }
+            if message.tool_calls:
+                assistant_message["tool_calls"] = [
+                    {
+                        "id": tool_call.id,
+                        "type": "function",
+                        "function": {
+                            "name": tool_call.function.name,
+                            "arguments": tool_call.function.arguments,
+                        },
+                    }
+                    for tool_call in message.tool_calls
+                ]
+            messages.append(assistant_message)
+            logger.info(f"Message count: {len(messages)}, Added assistant message with tool calls: {[tc.function.name for tc in message.tool_calls] if message.tool_calls else None}")
+
+            # If there are no tool calls, we're done
+            if not message.tool_calls:
+                return response
+
+            # Process all tool calls
+            tool_responses = []
+            for tool_call in message.tool_calls:
+                tool = self._tools.get(tool_call.function.name)
+                if not tool:
+                    raise ValueError(f"Unknown tool: {tool_call.function.name}")
+
+                # Parse arguments if they're a string
+                arguments = tool_call.function.arguments
+                if isinstance(arguments, str):
+                    try:
+                        arguments = json.loads(arguments)
+                    except json.JSONDecodeError as e:
+                        logger.error(f"Failed to parse tool arguments: {e}")
+                        raise ValueError(f"Invalid tool arguments format: {arguments}")
+
+                # Call the tool
+                logger.info(f"Calling tool: {tool.name} with args: {self._truncate_dict(arguments)}")
+                if inspect.iscoroutinefunction(tool.function):
+                    result = await tool.function(**arguments)
+                else:
+                    result = tool.function(**arguments)
+
+                # Convert result to JSON string if it's not already
+                if not isinstance(result, str):
+                    result = json.dumps(result)
+
+                # Create the tool response
+                tool_response = {
+                    "role": "tool",
+                    "tool_call_id": tool_call.id,
+                    "content": result,
+                }
+                tool_responses.append(tool_response)
+                logger.info(f"Created tool response for call ID {tool_call.id}")
+
+            # Add all tool responses to the messages
+            messages.extend(tool_responses)
+            logger.info(f"Message count: {len(messages)}, Added {len(tool_responses)} tool responses to messages")
+
+            # Validate the message sequence
+            self._validate_message_sequence(messages, validate_responses=True)
+
+            # Log the messages we're about to send
+            self._log_messages(messages, validate_responses=False)
+
+            # Continue the loop to get the next response
+            continue 

{mbxai-0.6.10 → mbxai-0.6.11}/uv.lock

@@ -292,11 +292,11 @@ wheels = [
 
 [[package]]
 name = "httpx-sse"
-version = "0.6.10"
+version = "0.6.11"
 source = { registry = "https://pypi.org/simple" }
-sdist = { url = "https://files.pythonhosted.org/packages/4c/60/8f4281fa9bbf3c8034fd54c0e7412e66edbab6bc74c4996bd616f8d0406e/httpx-sse-0.6.10.tar.gz", hash = "sha256:1e81a3a3070ce322add1d3529ed42eb5f70817f45ed6ec915ab753f961139721", size = 12624 }
+sdist = { url = "https://files.pythonhosted.org/packages/4c/60/8f4281fa9bbf3c8034fd54c0e7412e66edbab6bc74c4996bd616f8d0406e/httpx-sse-0.6.11.tar.gz", hash = "sha256:1e81a3a3070ce322add1d3529ed42eb5f70817f45ed6ec915ab753f961139721", size = 12624 }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/e1/9b/a181f281f65d776426002f330c31849b86b31fc9d848db62e16f03ff739f/httpx_sse-0.6.10-py3-none-any.whl", hash = "sha256:f329af6eae57eaa2bdfd962b42524764af68075ea87370a2de920af5341e318f", size = 7819 },
+    { url = "https://files.pythonhosted.org/packages/e1/9b/a181f281f65d776426002f330c31849b86b31fc9d848db62e16f03ff739f/httpx_sse-0.6.11-py3-none-any.whl", hash = "sha256:f329af6eae57eaa2bdfd962b42524764af68075ea87370a2de920af5341e318f", size = 7819 },
 ]
 
 [[package]]
@@ -446,7 +446,7 @@ wheels = [
 
 [[package]]
 name = "mbxai"
-version = "0.6.10"
+version = "0.6.11"
 source = { editable = "." }
 dependencies = [
     { name = "fastapi" },
@@ -980,14 +980,14 @@ wheels = [
 
 [[package]]
 name = "typing-inspection"
-version = "0.6.10"
+version = "0.6.11"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "typing-extensions" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/82/5c/e6082df02e215b846b4b8c0b887a64d7d08ffaba30605502639d44c06b82/typing_inspection-0.6.10.tar.gz", hash = "sha256:9765c87de36671694a67904bf2c96e395be9c6439bb6c87b5142569dcdd65122", size = 76222 }
+sdist = { url = "https://files.pythonhosted.org/packages/82/5c/e6082df02e215b846b4b8c0b887a64d7d08ffaba30605502639d44c06b82/typing_inspection-0.6.11.tar.gz", hash = "sha256:9765c87de36671694a67904bf2c96e395be9c6439bb6c87b5142569dcdd65122", size = 76222 }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/31/08/aa4fdfb71f7de5176385bd9e90852eaf6b5d622735020ad600f2bab54385/typing_inspection-0.6.10-py3-none-any.whl", hash = "sha256:50e72559fcd2a6367a19f7a7e610e6afcb9fac940c650290eed893d61386832f", size = 14125 },
+    { url = "https://files.pythonhosted.org/packages/31/08/aa4fdfb71f7de5176385bd9e90852eaf6b5d622735020ad600f2bab54385/typing_inspection-0.6.11-py3-none-any.whl", hash = "sha256:50e72559fcd2a6367a19f7a7e610e6afcb9fac940c650290eed893d61386832f", size = 14125 },
 ]
 
 [[package]]

mbxai-0.6.10/src/mbxai/tools/client.py

@@ -1,305 +0,0 @@
-"""
-ToolClient implementation for MBX AI.
-"""
-
-from typing import Any, Callable, TypeVar, cast
-import logging
-import inspect
-import json
-from pydantic import BaseModel
-from ..openrouter import OpenRouterClient
-from .types import Tool, ToolCall
-import asyncio
-
-logger = logging.getLogger(__name__)
-
-T = TypeVar("T", bound=BaseModel)
-
-class ToolClient:
-    """Base class for tool clients."""
-
-    def __init__(self, openrouter_client: OpenRouterClient):
-        """Initialize the tool client."""
-        self._openrouter_client = openrouter_client
-        self._tools: dict[str, Tool] = {}
-
-    def register_tool(self, tool: Tool) -> None:
-        """Register a tool."""
-        self._tools[tool.name] = tool
-
-    async def invoke_tool(self, tool_name: str, **kwargs: Any) -> Any:
-        """Invoke a tool by name."""
-        tool = self._tools.get(tool_name)
-        if not tool:
-            raise ValueError(f"Tool {tool_name} not found")
-        
-        if not tool.function:
-            raise ValueError(f"Tool {tool_name} has no function implementation")
-            
-        return await tool.function(**kwargs)
-
-    async def chat(self, messages: list[dict[str, str]], model: str | None = None) -> Any:
-        """Process a chat request with tools."""
-        # Convert tools to OpenAI function format
-        functions = [tool.to_openai_function() for tool in self._tools.values()]
-        
-        # Make the chat request
-        response = await self._openrouter_client.chat_completion(
-            messages=messages,
-            model=model,
-            functions=functions,
-        )
-        
-        # Validate response
-        if not response:
-            raise ValueError("No response received from OpenRouter")
-            
-        if not response.choices:
-            raise ValueError("Response missing choices")
-            
-        choice = response.choices[0]
-        if not choice:
-            raise ValueError("Empty choice in response")
-            
-        message = choice.message
-        if not message:
-            raise ValueError("Choice missing message")
-            
-        # If message has function call, execute it
-        if message.function_call:
-            tool_name = message.function_call.name
-            tool_args = json.loads(message.function_call.arguments)
-            
-            # Invoke the tool
-            tool_response = await self.invoke_tool(tool_name, **tool_args)
-            
-            # Add tool response to messages
-            messages.append({
-                "role": "assistant",
-                "content": None,
-                "function_call": {
-                    "name": tool_name,
-                    "arguments": message.function_call.arguments,
-                },
-            })
-            messages.append({
-                "role": "function",
-                "name": tool_name,
-                "content": json.dumps(tool_response),
-            })
-            
-            # Get final response
-            final_response = await self._openrouter_client.chat_completion(
-                messages=messages,
-                model=model,
-            )
-            
-            if not final_response or not final_response.choices:
-                raise ValueError("No response received after tool execution")
-                
-            return final_response
-            
-        return response
-
-    def _truncate_content(self, content: str | None, max_length: int = 100) -> str:
-        """Truncate content for logging."""
-        if not content:
-            return "None"
-        if len(content) <= max_length:
-            return content
-        return content[:max_length] + "..."
-
-    def _truncate_dict(self, data: dict[str, Any], max_length: int = 50) -> str:
-        """Truncate dictionary values for logging."""
-        if not data:
-            return "{}"
-        truncated = {}
-        for k, v in data.items():
-            if isinstance(v, str):
-                truncated[k] = self._truncate_content(v, max_length)
-            elif isinstance(v, dict):
-                truncated[k] = self._truncate_dict(v, max_length)
-            else:
-                truncated[k] = str(v)[:max_length] + "..." if len(str(v)) > max_length else v
-        return str(truncated)
-
-    def _validate_message_sequence(self, messages: list[dict[str, Any]], validate_responses: bool = True) -> None:
-        """Validate the message sequence for tool calls and responses.
-        
-        Args:
-            messages: The message sequence to validate
-            validate_responses: Whether to validate that all tool calls have responses
-        """
-        tool_call_ids = set()
-        tool_response_ids = set()
-        
-        for i, msg in enumerate(messages):
-            role = msg.get("role")
-            if role == "assistant" and "tool_calls" in msg:
-                # Track tool calls
-                for tc in msg["tool_calls"]:
-                    tool_call_ids.add(tc["id"])
-                    logger.info(f"Found tool call {tc['id']} for {tc['function']['name']} in message {i}")
-            elif role == "tool":
-                # Track tool responses
-                tool_response_ids.add(msg["tool_call_id"])
-                logger.info(f"Found tool response for call ID {msg['tool_call_id']} in message {i}")
-        
-        # Only validate responses if requested
-        if validate_responses:
-            # Check for missing responses
-            missing_responses = tool_call_ids - tool_response_ids
-            if missing_responses:
-                logger.error(f"Missing tool responses for call IDs: {missing_responses}")
-                logger.error("Message sequence:")
-                for i, msg in enumerate(messages):
-                    role = msg.get("role", "unknown")
-                    if role == "assistant" and "tool_calls" in msg:
-                        logger.error(f"  Message {i} - Assistant with tool calls: {[tc['id'] for tc in msg['tool_calls']]}")
-                    elif role == "tool":
-                        logger.error(f"  Message {i} - Tool response for call ID: {msg['tool_call_id']}")
-                    else:
-                        logger.error(f"  Message {i} - {role}: {self._truncate_content(msg.get('content'))}")
-                raise ValueError(f"Invalid message sequence: missing responses for tool calls {missing_responses}")
-
-    def _log_messages(self, messages: list[dict[str, Any]], validate_responses: bool = True) -> None:
-        """Log the messages being sent to OpenRouter.
-        
-        Args:
-            messages: The messages to log
-            validate_responses: Whether to validate that all tool calls have responses
-        """
-        logger.info("Sending messages to OpenRouter:")
-        for i, msg in enumerate(messages):
-            role = msg.get("role", "unknown")
-            content = self._truncate_content(msg.get("content"))
-            tool_calls = msg.get("tool_calls", [])
-            tool_call_id = msg.get("tool_call_id")
-            
-            if tool_calls:
-                tool_call_info = [
-                    f"{tc['function']['name']}(id={tc['id']})"
-                    for tc in tool_calls
-                ]
-                logger.info(f"  Message {i} - {role}: content='{content}', tool_calls={tool_call_info}")
-            elif tool_call_id:
-                logger.info(f"  Message {i} - {role}: content='{content}', tool_call_id={tool_call_id}")
-            else:
-                logger.info(f"  Message {i} - {role}: content='{content}'")
-        
-        # Validate message sequence
-        self._validate_message_sequence(messages, validate_responses)
-
-    async def _process_tool_calls(self, message: Any, messages: list[dict[str, Any]]) -> None:
-        """Process all tool calls in a message.
-        
-        Args:
-            message: The message containing tool calls
-            messages: The list of messages to add responses to
-        """
-        if not message.tool_calls:
-            return
-
-        # Process all tool calls first
-        tool_responses = []
-        for tool_call in message.tool_calls:
-            tool = self._tools.get(tool_call.function.name)
-            if not tool:
-                raise ValueError(f"Unknown tool: {tool_call.function.name}")
-
-            # Parse arguments if they're a string
-            arguments = tool_call.function.arguments
-            if isinstance(arguments, str):
-                try:
-                    arguments = json.loads(arguments)
-                except json.JSONDecodeError as e:
-                    logger.error(f"Failed to parse tool arguments: {e}")
-                    raise ValueError(f"Invalid tool arguments format: {arguments}")
-
-            # Call the tool
-            logger.info(f"Calling tool: {tool.name} with args: {self._truncate_dict(arguments)}")
-            if inspect.iscoroutinefunction(tool.function):
-                result = await tool.function(**arguments)
-            else:
-                result = tool.function(**arguments)
-
-            # Convert result to JSON string if it's not already
-            if not isinstance(result, str):
-                result = json.dumps(result)
-
-            # Create the tool response
-            tool_response = {
-                "role": "tool",
-                "tool_call_id": tool_call.id,
-                "content": result,
-            }
-            tool_responses.append(tool_response)
-            logger.info(f"Created tool response for call ID {tool_call.id}")
-
-        # Add all tool responses to the messages
-        messages.extend(tool_responses)
-        logger.info(f"Message count: {len(messages)}, Added {len(tool_responses)} tool responses to messages")
-
-        # Validate the message sequence
-        self._validate_message_sequence(messages, validate_responses=True)
-
-        # Log the messages we're about to send
-        self._log_messages(messages, validate_responses=False)
-
-    async def parse(
-        self,
-        messages: list[dict[str, str]],
-        response_format: type[T],
-        *,
-        model: str | None = None,
-        **kwargs: Any,
-    ) -> Any:
-        """Chat with the model and parse the response into a Pydantic model.
-
-        Args:
-            messages: The conversation messages
-            response_format: The Pydantic model to parse the response into
-            model: Optional model override
-            **kwargs: Additional parameters for the chat completion
-
-        Returns:
-            The parsed response from the model
-        """
-        # First use our own chat function to handle any tool calls
-        response = await self.chat(
-            messages=messages,
-            model=model,
-            **kwargs,
-        )
-
-        if not response or not response.choices:
-            raise ValueError("No response received from OpenRouter")
-            
-        choice = response.choices[0]
-        if not choice:
-            raise ValueError("Empty choice in response")
-            
-        message = choice.message
-        if not message:
-            raise ValueError("Choice missing message")
-
-        # If we still have tool calls, something went wrong
-        if message.tool_calls:
-            raise ValueError("Unexpected tool calls in final response")
-
-        # Ensure we have content to parse
-        if not message.content:
-            raise ValueError("No content in final response to parse")
-
-        # Now that we have the final response, parse it into the desired format
-        try:
-            final_response = await self._openrouter_client.chat_completion_parse(
-                messages=messages,
-                response_format=response_format,
-                model=model,
-                **kwargs,
-            )
-            return final_response
-        except Exception as e:
-            logger.error(f"Failed to parse response: {e}")
-            raise ValueError(f"Failed to parse response as {response_format.__name__}: {str(e)}")