metorial-openai-compatible 1.0.0rc1

OpenAI-compatible provider base for Metorial

metorial-openai-compatible

Base package for OpenAI-compatible provider integrations for Metorial. This package provides shared functionality for providers that use OpenAI's function calling format.

Installation

pip install metorial-openai-compatible
# or
uv add metorial-openai-compatible
# or
poetry add metorial-openai-compatible

Features

• 🔧 OpenAI Format: Standard OpenAI function calling format
• ✅ Strict Mode: Configurable strict parameter validation
• 📡 Session Management: Automatic tool lifecycle handling
• 🔄 Format Conversion: Converts Metorial tools to OpenAI function format
• ⚡ Async Support: Full async/await support
• 🏗️ Base Class: Foundation for provider-specific implementations

Usage

Direct Usage

import asyncio
from metorial import Metorial
from metorial_openai_compatible import MetorialOpenAICompatibleSession

async def main():
    # Initialize Metorial
    metorial = Metorial(api_key="your-metorial-api-key")
    
    # Create session with your server deployments
    async with metorial.session(["your-server-deployment-id"]) as session:
        # Create OpenAI-compatible wrapper
        openai_session = MetorialOpenAICompatibleSession(
            session.tool_manager,
            with_strict=True  # Enable strict mode
        )
        
        # Use with any OpenAI-compatible client
        tools = openai_session.tools
        
        # Handle tool calls from response
        tool_responses = await openai_session.call_tools(tool_calls)

asyncio.run(main())

As Base Class

This package is primarily used as a base for provider-specific packages:

from metorial_openai_compatible import MetorialOpenAICompatibleSession

class MyProviderSession(MetorialOpenAICompatibleSession):
    def __init__(self, tool_mgr):
        # Configure strict mode based on provider capabilities
        super().__init__(tool_mgr, with_strict=False)

Using Convenience Functions

from metorial_openai_compatible import build_openai_compatible_tools, call_openai_compatible_tools

async def example_with_functions():
    # Get tools in OpenAI format
    tools = build_openai_compatible_tools(tool_manager, with_strict=True)
    
    # Call tools from OpenAI-compatible response
    tool_messages = await call_openai_compatible_tools(tool_manager, tool_calls)

API Reference

MetorialOpenAICompatibleSession

Main session class for OpenAI-compatible integration.

session = MetorialOpenAICompatibleSession(tool_manager, with_strict=False)

Parameters:

• tool_manager: Metorial tool manager instance
• with_strict: Enable strict parameter validation (default: False)

Properties:

• tools: List of tools in OpenAI function calling format

Methods:

• async call_tools(tool_calls): Execute tool calls and return tool messages

build_openai_compatible_tools(tool_mgr, with_strict=False)

Build OpenAI-compatible tool definitions.

Parameters:

• tool_mgr: Tool manager instance
• with_strict: Enable strict mode (default: False)

Returns: List of tool definitions in OpenAI format

call_openai_compatible_tools(tool_mgr, tool_calls)

Execute tool calls from OpenAI-compatible response.

Returns: List of tool messages

Tool Format

Tools are converted to OpenAI's function calling format:

{
    "type": "function",
    "function": {
        "name": "tool_name",
        "description": "Tool description",
        "parameters": {
            "type": "object",
            "properties": {...},
            "required": [...]
        },
        "strict": True  # Only if with_strict=True
    }
}

Strict Mode

When with_strict=True, the strict field is added to function definitions for providers that support strict parameter validation (like OpenAI and XAI).

Provider Implementations

This package serves as the base for:

• metorial-xai: XAI (Grok) with strict mode enabled
• metorial-deepseek: DeepSeek without strict mode
• metorial-togetherai: Together AI without strict mode

Error Handling

try:
    tool_messages = await session.call_tools(tool_calls)
except Exception as e:
    print(f"Tool execution failed: {e}")

Tool errors are returned as tool messages with error content.

Dependencies

• metorial-mcp-session>=1.0.0
• typing-extensions>=4.0.0

License

MIT License - see LICENSE file for details.