# 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
```bash
pip install metorial-openai-compatible
# or
uv add metorial-openai-compatible
# or
poetry add metorial-openai-compatible
```
## Features
- š§ **OpenAI Format**: Standard OpenAI function calling format
- š” **Session Management**: Automatic tool lifecycle handling
- š **Format Conversion**: Converts Metorial tools to OpenAI function format
- ā” **Async Support**: Full async/await support
## Usage
### Quick Start (Recommended)
This package serves as a base for provider-specific implementations. For end-user usage, use the specific provider packages like `metorial-xai`, `metorial-deepseek`, or `metorial-togetherai`.
### Direct Usage (Advanced)
```python
import asyncio
from openai import AsyncOpenAI
from metorial import Metorial
from metorial_openai_compatible import MetorialOpenAICompatibleSession
async def main():
# Initialize clients
metorial = Metorial(api_key="...your-metorial-api-key...") # async by default
compatible_client = AsyncOpenAI(
api_key="...your-provider-api-key...",
base_url="https://your-provider-url/v1"
)
# Run with automatic session management
response = await metorial.run(
"What are the latest commits in the metorial/websocket-explorer repository?",
"...your-mcp-server-deployment-id...", # can also be list
compatible_client,
model="your-model-name",
max_iterations=25
)
print("Response:", response)
asyncio.run(main())
```
### Streaming Chat
```python
import asyncio
from openai import AsyncOpenAI
from metorial import Metorial
from metorial.types import StreamEventType
async def example():
# Initialize clients
metorial = Metorial(api_key="...your-metorial-api-key...")
compatible_client = AsyncOpenAI(
api_key="...your-provider-api-key...",
base_url="https://your-provider-url/v1"
)
# Streaming chat with real-time responses
async def stream_action(session):
messages = [
{"role": "user", "content": "Explain quantum computing"}
]
async for event in metorial.stream(
compatible_client, session, messages,
model="your-model-name",
max_iterations=25
):
if event.type == StreamEventType.CONTENT:
print(f"š¤ {event.content}", end="", flush=True)
elif event.type == StreamEventType.TOOL_CALL:
print(f"\nš§ Executing {len(event.tool_calls)} tool(s)...")
elif event.type == StreamEventType.COMPLETE:
print(f"\nā
Complete!")
await metorial.with_session("...your-server-deployment-id...", stream_action)
asyncio.run(example())
```
### Advanced Usage with Session Management
```python
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:
```python
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
```python
from metorial_openai_compatible import build_openai_compatible_tools, call_openai_compatible_tools
async def example():
# 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.
```python
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:
```python
{
"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
```python
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.
## License
MIT License - see [LICENSE](../../LICENSE) file for details.
Raw data
{
"_id": null,
"home_page": null,
"name": "metorial-openai-compatible",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.10",
"maintainer_email": null,
"keywords": "ai, compatible, llm, metorial, openai",
"author": null,
"author_email": "Metorial Team <support@metorial.com>",
"download_url": "https://files.pythonhosted.org/packages/34/2f/c407ce8969e60bdf44806e7e27b46c4eecda86752cfce4c364ed4e282664/metorial_openai_compatible-1.0.4.tar.gz",
"platform": null,
"description": "# metorial-openai-compatible\n\nBase package for OpenAI-compatible provider integrations for Metorial. This package provides shared functionality for providers that use OpenAI's function calling format.\n\n## Installation\n\n```bash\npip install metorial-openai-compatible\n# or\nuv add metorial-openai-compatible\n# or\npoetry add metorial-openai-compatible\n```\n\n## Features\n\n- \ud83d\udd27 **OpenAI Format**: Standard OpenAI function calling format\n- \ud83d\udce1 **Session Management**: Automatic tool lifecycle handling\n- \ud83d\udd04 **Format Conversion**: Converts Metorial tools to OpenAI function format\n- \u26a1 **Async Support**: Full async/await support\n\n## Usage\n\n### Quick Start (Recommended)\n\nThis package serves as a base for provider-specific implementations. For end-user usage, use the specific provider packages like `metorial-xai`, `metorial-deepseek`, or `metorial-togetherai`.\n\n### Direct Usage (Advanced)\n\n```python\nimport asyncio\nfrom openai import AsyncOpenAI\nfrom metorial import Metorial\nfrom metorial_openai_compatible import MetorialOpenAICompatibleSession\n\nasync def main():\n # Initialize clients\n metorial = Metorial(api_key=\"...your-metorial-api-key...\") # async by default\n compatible_client = AsyncOpenAI(\n api_key=\"...your-provider-api-key...\", \n base_url=\"https://your-provider-url/v1\"\n )\n \n # Run with automatic session management\n response = await metorial.run(\n \"What are the latest commits in the metorial/websocket-explorer repository?\",\n \"...your-mcp-server-deployment-id...\", # can also be list\n compatible_client,\n model=\"your-model-name\",\n max_iterations=25\n )\n \n print(\"Response:\", response)\n\nasyncio.run(main())\n```\n\n### Streaming Chat\n\n```python\nimport asyncio\nfrom openai import AsyncOpenAI\nfrom metorial import Metorial\nfrom metorial.types import StreamEventType\n\nasync def example():\n # Initialize clients\n metorial = Metorial(api_key=\"...your-metorial-api-key...\")\n compatible_client = AsyncOpenAI(\n api_key=\"...your-provider-api-key...\",\n base_url=\"https://your-provider-url/v1\"\n )\n \n # Streaming chat with real-time responses\n async def stream_action(session):\n messages = [\n {\"role\": \"user\", \"content\": \"Explain quantum computing\"}\n ]\n \n async for event in metorial.stream(\n compatible_client, session, messages, \n model=\"your-model-name\",\n max_iterations=25\n ):\n if event.type == StreamEventType.CONTENT:\n print(f\"\ud83e\udd16 {event.content}\", end=\"\", flush=True)\n elif event.type == StreamEventType.TOOL_CALL:\n print(f\"\\n\ud83d\udd27 Executing {len(event.tool_calls)} tool(s)...\")\n elif event.type == StreamEventType.COMPLETE:\n print(f\"\\n\u2705 Complete!\")\n \n await metorial.with_session(\"...your-server-deployment-id...\", stream_action)\n\nasyncio.run(example())\n```\n\n### Advanced Usage with Session Management\n\n```python\nimport asyncio\nfrom metorial import Metorial\nfrom metorial_openai_compatible import MetorialOpenAICompatibleSession\n\nasync def main():\n # Initialize Metorial\n metorial = Metorial(api_key=\"...your-metorial-api-key...\")\n \n # Create session with your server deployments\n async with metorial.session([\"...your-server-deployment-id...\"]) as session:\n # Create OpenAI-compatible wrapper\n openai_session = MetorialOpenAICompatibleSession(\n session.tool_manager,\n with_strict=True # Enable strict mode\n )\n \n # Use with any OpenAI-compatible client\n tools = openai_session.tools\n \n # Handle tool calls from response\n tool_responses = await openai_session.call_tools(tool_calls)\n\nasyncio.run(main())\n```\n\n### As Base Class\n\nThis package is primarily used as a base for provider-specific packages:\n\n```python\nfrom metorial_openai_compatible import MetorialOpenAICompatibleSession\n\nclass MyProviderSession(MetorialOpenAICompatibleSession):\n def __init__(self, tool_mgr):\n # Configure strict mode based on provider capabilities\n super().__init__(tool_mgr, with_strict=False)\n```\n\n### Using Convenience Functions\n\n```python\nfrom metorial_openai_compatible import build_openai_compatible_tools, call_openai_compatible_tools\n\nasync def example():\n # Get tools in OpenAI format\n tools = build_openai_compatible_tools(tool_manager, with_strict=True)\n \n # Call tools from OpenAI-compatible response\n tool_messages = await call_openai_compatible_tools(tool_manager, tool_calls)\n```\n\n## API Reference\n\n### `MetorialOpenAICompatibleSession`\n\nMain session class for OpenAI-compatible integration.\n\n```python\nsession = MetorialOpenAICompatibleSession(tool_manager, with_strict=False)\n```\n\n**Parameters:**\n- `tool_manager`: Metorial tool manager instance\n- `with_strict`: Enable strict parameter validation (default: False)\n\n**Properties:**\n- `tools`: List of tools in OpenAI function calling format\n\n**Methods:**\n- `async call_tools(tool_calls)`: Execute tool calls and return tool messages\n\n### `build_openai_compatible_tools(tool_mgr, with_strict=False)`\n\nBuild OpenAI-compatible tool definitions.\n\n**Parameters:**\n- `tool_mgr`: Tool manager instance\n- `with_strict`: Enable strict mode (default: False)\n\n**Returns:** List of tool definitions in OpenAI format\n\n### `call_openai_compatible_tools(tool_mgr, tool_calls)`\n\nExecute tool calls from OpenAI-compatible response.\n\n**Returns:** List of tool messages\n\n## Tool Format\n\nTools are converted to OpenAI's function calling format:\n\n```python\n{\n \"type\": \"function\",\n \"function\": {\n \"name\": \"tool_name\",\n \"description\": \"Tool description\",\n \"parameters\": {\n \"type\": \"object\",\n \"properties\": {...},\n \"required\": [...]\n },\n \"strict\": True # Only if with_strict=True\n }\n}\n```\n\n## Strict Mode\n\nWhen `with_strict=True`, the `strict` field is added to function definitions for providers that support strict parameter validation (like OpenAI and XAI).\n\n## Provider Implementations\n\nThis package serves as the base for:\n\n- **metorial-xai**: XAI (Grok) with strict mode enabled\n- **metorial-deepseek**: DeepSeek without strict mode\n- **metorial-togetherai**: Together AI without strict mode\n\n## Error Handling\n\n```python\ntry:\n tool_messages = await session.call_tools(tool_calls)\nexcept Exception as e:\n print(f\"Tool execution failed: {e}\")\n```\n\nTool errors are returned as tool messages with error content.\n\n## License\n\nMIT License - see [LICENSE](../../LICENSE) file for details.\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "OpenAI-compatible provider base for Metorial",
"version": "1.0.4",
"project_urls": {
"Documentation": "https://metorial.com/docs",
"Homepage": "https://metorial.com",
"Repository": "https://github.com/metorial/metorial-python"
},
"split_keywords": [
"ai",
" compatible",
" llm",
" metorial",
" openai"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "7acf253e0bdb8daa2ffa89e76fe1c590eb01a9d78d2770a9498fe74b566f32cd",
"md5": "bfc321fa5dbca10b5eaa1181abaa0306",
"sha256": "b006213fd2edfce55db35b23df6bb02752d3b71be862ef4189626fca135e671f"
},
"downloads": -1,
"filename": "metorial_openai_compatible-1.0.4-py3-none-any.whl",
"has_sig": false,
"md5_digest": "bfc321fa5dbca10b5eaa1181abaa0306",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.10",
"size": 5991,
"upload_time": "2025-10-30T05:03:15",
"upload_time_iso_8601": "2025-10-30T05:03:15.307396Z",
"url": "https://files.pythonhosted.org/packages/7a/cf/253e0bdb8daa2ffa89e76fe1c590eb01a9d78d2770a9498fe74b566f32cd/metorial_openai_compatible-1.0.4-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "342fc407ce8969e60bdf44806e7e27b46c4eecda86752cfce4c364ed4e282664",
"md5": "20f1f39ccabe7e4e3dad193f675cf531",
"sha256": "def1a94dcd1429612983a37388bd3f198ff227da3444b5c1870b8ec2d6a74bf1"
},
"downloads": -1,
"filename": "metorial_openai_compatible-1.0.4.tar.gz",
"has_sig": false,
"md5_digest": "20f1f39ccabe7e4e3dad193f675cf531",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.10",
"size": 6869,
"upload_time": "2025-10-30T05:03:24",
"upload_time_iso_8601": "2025-10-30T05:03:24.525092Z",
"url": "https://files.pythonhosted.org/packages/34/2f/c407ce8969e60bdf44806e7e27b46c4eecda86752cfce4c364ed4e282664/metorial_openai_compatible-1.0.4.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-10-30 05:03:24",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "metorial",
"github_project": "metorial-python",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"lcname": "metorial-openai-compatible"
}
JSON
