metorial-core 1.0.9

Core components for Metorial Python SDK

Metorial Python SDK

The official Python SDK for Metorial.

Available Providers

Provider	Format	Description
OpenAI	OpenAI function calling	GPT-4, GPT-3.5, etc.
Anthropic	Claude tool format	Claude 3.5, Claude 3, etc.
Google	Gemini function declarations	Gemini Pro, Gemini Flash
Mistral	Mistral function calling	Mistral Large, Codestral
DeepSeek	OpenAI-compatible	DeepSeek Chat, DeepSeek Coder
TogetherAI	OpenAI-compatible	Llama, Mixtral, etc.
XAI	OpenAI-compatible	Grok models
AI SDK	Framework tools	Vercel AI SDK, etc.

Installation

# Install core metorial package (includes all provider adapters)
pip install metorial

# Install with specific providers (includes provider client libraries)
pip install metorial[openai,anthropic,google,mistral,deepseek,togetherai,xai]

Quick Start

Simple Usage

import asyncio
from metorial import Metorial
from openai import AsyncOpenAI

async def main():
  metorial = Metorial(api_key="your-metorial-api-key")
  openai = AsyncOpenAI(api_key="your-openai-api-key")
  
  response = await metorial.run(
    message="Search Hackernews for the latest AI discussions.",
    server_deployments=["hacker-news-server-deployment"],
    client=openai,
    model="gpt-4o",
    max_steps=25    # optional
  )
  
  print("Response:", response.text)

asyncio.run(main())

💡 Tip for Jupyter/Colab Users: If you're running in a Jupyter notebook or Google Colab, you can skip the async def main(): wrapper and asyncio.run() and just use await directly at the top level.

OAuth + Multiple Deployments

For integrations requiring OAuth authentication (like Google Calendar) and multiple server deployments:

import asyncio
import os
from metorial import Metorial
from anthropic import AsyncAnthropic

async def main():
  metorial = Metorial(api_key=os.getenv("METORIAL_API_KEY"))
  anthropic = AsyncAnthropic(api_key=os.getenv("ANTHROPIC_API_KEY"))

  # Create OAuth session for authenticated services
  google_cal_deployment_id = os.getenv("GOOGLE_CALENDAR_DEPLOYMENT_ID")
  
  print("🔗 Creating OAuth session...")
  oauth_session = metorial.oauth.sessions.create(
    server_deployment_id=google_cal_deployment_id
  )

  print("OAuth URLs for user authentication:")
  print(f"   Google Calendar: {oauth_session.url}")

  print("\n⏳ Waiting for OAuth completion...")
  await metorial.oauth.wait_for_completion([oauth_session])
  print("✅ OAuth session completed!")

  # Use multiple server deployments with mixed auth
  hackernews_deployment_id = os.getenv("HACKERNEWS_DEPLOYMENT_ID")
  
  result = await metorial.run(
    message="""Search Hackernews for the latest AI discussions using the available tools. 
    Then create a calendar event using Google Calendar tools with [e-mail] for tomorrow at 2pm to discuss AI trends.""",
    server_deployments=[
      { "serverDeploymentId": google_cal_deployment_id, "oauthSessionId": oauth_session.id },
      { "serverDeploymentId": hackernews_deployment_id },
    ],
    client=anthropic,
    model="claude-sonnet-4-20250514",
    max_tokens=4096,
    max_steps=25,
  )
  print(result.text)

asyncio.run(main())

That's it! metorial.run() automatically:

• Creates a session with your MCP server
• Formats tools for your AI provider
• Handles the execution loop
• Manages tool execution
• Returns the final response

Advanced Usage with Provider Sessions

For more control over the conversation flow, you can use with_provider_session:

import asyncio
from metorial import Metorial, MetorialOpenAI
from openai import AsyncOpenAI

async def main():
  metorial = Metorial(api_key="your-metorial-api-key")
  openai = AsyncOpenAI(api_key="your-openai-api-key")

  messages = [
    {"role": "user", "content": "What are the top hackernews posts?"}
  ]

  async def session_action(session):
    for i in range(10):
      response = await openai.chat.completions.create(
        messages=messages,
        model="gpt-4o",
        tools=session["tools"]
      )

      choice = response.choices[0]
      tool_calls = choice.message.tool_calls

      if not tool_calls:
        print(choice.message.content)
        return

      # Execute tools through Metorial
      tool_responses = await session["callTools"](tool_calls)

      # Add to conversation
      messages.append({
        "role": "assistant", 
        "tool_calls": choice.message.tool_calls
      })
      messages.extend(tool_responses)

  await metorial.with_provider_session(
    MetorialOpenAI.chat_completions,
    [{"serverDeploymentId": "your-deployment-id"}],
    session_action
  )

asyncio.run(main())

This approach gives you full control over the conversation loop while still benefiting from Metorial's tool management.

Provider Examples

Metorial works with all major AI providers. Here are examples using metorial.run():

Example OpenAI (GPT-4, GPT-3.5)

from metorial import Metorial
from openai import AsyncOpenAI

metorial = Metorial(api_key="your-metorial-api-key")
openai = AsyncOpenAI(api_key="your-openai-api-key")

response = await metorial.run(
  message="What are the latest commits?",
  server_deployments=["your-deployment-id"],
  client=openai,
  model="gpt-4o"
)

Anthropic (Claude)

from metorial import Metorial
import anthropic

metorial = Metorial(api_key="your-metorial-api-key")
anthropic = anthropic.AsyncAnthropic(api_key="your-anthropic-api-key")

response = await metorial.run(
  message="What are the latest commits?",
  server_deployments=["your-deployment-id"],
  client=anthropic,
  model="claude-3-5-sonnet-20241022"
)

Google (Gemini)

from metorial import Metorial
import google.generativeai as genai

metorial = Metorial(api_key="your-metorial-api-key")
genai.configure(api_key="your-google-api-key")
google = genai.GenerativeModel('gemini-pro')

response = await metorial.run(
  message="What are the latest commits?",
  server_deployments=["your-deployment-id"],
  client=google,
  model="gemini-pro"
)

Mistral AI

from metorial import Metorial
from mistralai import AsyncMistral

metorial = Metorial(api_key="your-metorial-api-key")
mistral = AsyncMistral(api_key="your-mistral-api-key")

response = await metorial.run(
  message="What are the latest commits?",
  server_deployments=["your-deployment-id"],
  client=mistral,
  model="mistral-large-latest"
)

DeepSeek

from metorial import Metorial
from openai import AsyncOpenAI

metorial = Metorial(api_key="your-metorial-api-key")
deepseek = AsyncOpenAI(
  api_key="your-deepseek-api-key",
  base_url="https://api.deepseek.com"
)

response = await metorial.run(
  message="What are the latest commits?",
  server_deployments=["your-deployment-id"],
  client=deepseek,
  model="deepseek-chat"
)

Together AI

from metorial import Metorial
from openai import AsyncOpenAI

metorial = Metorial(api_key="your-metorial-api-key")
together = AsyncOpenAI(
  api_key="your-together-api-key",
  base_url="https://api.together.xyz/v1"
)

response = await metorial.run(
  message="What are the latest commits?",
  server_deployments=["your-deployment-id"],
  client=together,
  model="meta-llama/Llama-2-70b-chat-hf"
)

XAI (Grok)

from metorial import Metorial
from openai import AsyncOpenAI

metorial = Metorial(api_key="your-metorial-api-key")
xai = AsyncOpenAI(
  api_key="your-xai-api-key",
  base_url="https://api.x.ai/v1"
)

response = await metorial.run(
  message="What are the latest commits?",
  server_deployments=["your-deployment-id"],
  client=xai,
  model="grok-beta"
)

Error Handling

from metorial import MetorialAPIError

try:
  response = await metorial.run(
    message="What are the latest commits?",
    server_deployments=["your-deployment-id"],
    client=openai,
    model="gpt-4o"
  )
except MetorialAPIError as e:
  print(f"API Error: {e.message} (Status: {e.status_code})")
except Exception as e:
  print(f"Unexpected error: {e}")

Examples

Check out the examples/ directory for more comprehensive examples.

License

MIT License - see LICENSE file for details.

Support

• 📖 Documentation
• 🐛 GitHub Issues
• 📧 Email Support