toolflow 0.1.1

A Python library for Auto LLM tool calling with decorator-based function registration

Toolflow

A Python library that makes LLM tool calling as simple as decorating a function. Just wrap your OpenAI client and pass decorated functions directly to the tools parameter - no complex setup required.

import toolflow
from openai import OpenAI

client = toolflow.from_openai(OpenAI())

@toolflow.tool  
def get_weather(city: str) -> str:
    return f"Weather in {city}: Sunny, 72°F"

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Weather in NYC?"}],
    tools=[get_weather],  # Just pass your function! We'll handle the rest.
    max_tool_calls=5, # Maximum number of tool calls to execute as a safety measure.
)

print(response.choices[0].message.content)

Features

• 🎯 Simple decorator-based tool registration - Just use @tool to register functions
• 🔧 Multiple LLM provider support(Coming soon) - OpenAI, Anthropic (Claude), and extensible for others
• 📝 Automatic schema generation - Function signatures are automatically converted to JSON schemas
• ⚡ Automatic execution - Tools can be automatically executed when called by the LLM
• 🔄 Asynchronous support(Coming soon) - Tools can be called asynchronously
• 🔄 Streaming support(Coming soon) - Tools can be streamed

Installation

pip install toolflow

Quick Start

import os
import toolflow
from openai import OpenAI

# Create a toolflow wrapped client
client = toolflow.from_openai(OpenAI(api_key=os.getenv("OPENAI_API_KEY")))

# Define tools using the @tool decorator
@toolflow.tool
def get_weather(city: str) -> str:
    """Get the current weather for a city."""
    return f"Weather in {city}: Sunny, 72°F"

@toolflow.tool
def add_numbers(a: int, b: int) -> int:
    """Add two numbers and return the result."""
    return a + b

# Use the familiar OpenAI API with your functions as tools
response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {"role": "user", "content": "What's the weather in San Francisco? Also, what's 15 + 27?"}
    ],
    tools=[get_weather, add_numbers]
)

print("Response:", response.choices[0].message.content)

Core Concepts

1. Decorating Functions with @tool

The @tool decorator adds metadata to functions so they can be used as LLM tools:

import toolflow

@toolflow.tool
def search_database(query: str, limit: int = 10) -> list:
    """Search the database for matching records."""
    # Your implementation here
    return ["result1", "result2"]

# The function can now be passed directly to the tools parameter

2. Custom Tool Names and Descriptions

@toolflow.tool(name="db_search", description="Search our product database")
def search_products(query: str) -> list:
    """Original docstring."""
    return search_results

3. Multiple LLM Providers

# OpenAI
openai_client = toolflow.from_openai(OpenAI(api_key="your-openai-key"))

# Anthropic Claude
claude_client = toolflow.from_anthropic(anthropic.Anthropic(api_key="your-anthropic-key"))

4. Direct Function Calls

# You can still call your functions directly
@toolflow.tool
def get_weather(city: str) -> str:
    return f"Weather in {city}: Sunny"

# Direct call
result = get_weather("New York")
print(result)  # "Weather in New York: Sunny"

API Reference

@tool decorator

@toolflow.tool(name=None, description=None)

• name: Custom tool name (defaults to function name)
• description: Tool description (defaults to docstring)

from_openai(client)

Wraps an OpenAI client to support toolflow functions:

import openai
import toolflow

client = toolflow.from_openai(openai.OpenAI())

Enhanced chat.completions.create()

When using a wrapped client, the create method gains additional parameters:

• tools: List of toolflow decorated functions or regular tool dicts
• max_tool_calls: Maximum number of tool calls to execute (default: 5)

Tool Function Methods

Every @tool decorated function gets these methods:

• _tool_metadata - JSON schema for the tool

Development

# Install development dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Format code
black toolflow/
isort toolflow/

# Type checking
mypy toolflow/

Contributing

1. Fork the repository
2. Create a feature branch
3. Add tests for new functionality
4. Ensure all tests pass
5. Submit a pull request

License

MIT License - see LICENSE file for details.