eleventools 0.1.3

Use your Python functions as ElevenLabs agent tools

ElevenTools

Take a shortcut in working with ElevenLabs agents by using your local Python functions as conversational AI tools.

Overview

ElevenTools is a Python library that provides a convenient path to use your local functions with ElevenLabs conversational AI agents. Instead of manually designing endpoints, using the UI or CLI to create/update agent configurations and tools, or assigning tools through the dashboard, this library allows you to:

• Define tools with simple Python decorators - Turn any function into an ElevenLabs agent tool
• Automatically sync with ElevenLabs API - Creates and updates tools seamlessly
• Skip manual configuration - No need to manually design webhooks or configure the dashboard
• Focus on business logic - Write your tool functionality, let ElevenTools handle the integration

How It Works

1. Agent-Specific Tool Management: Each WebhookToolset instance is tied to a specific ElevenLabs agent
2. Automatic Tool Naming: Tools are created with the pattern {agentName}_{toolName} to avoid conflicts
3. Intelligent Sync: Only creates new tools or updates existing ones when changes are detected
4. Automatic Assignment: Tools are automatically assigned to the specified agent after sync

Quick Start

Installation

pip install eleventools

Basic Usage

from eleventools import WebhookToolset
import asyncio

# Initialize for a specific agent
toolset = WebhookToolset(
    base_url="https://your-webhook-server.com",
    xi_api_key="your_elevenlabs_api_key", 
    agent_id="your_agent_id",
    agent_name="your_agent_name"
)

# Define tools using the @toolset.tool() decorator
@toolset.tool(description="Get current weather for any city")
def get_weather(city: str, units: str = "celsius"):
    """Get weather information for a city."""
    # Your weather API logic here
    return {"weather": f"Sunny in {city}", "temperature": "22°C"}

@toolset.tool(description="Perform mathematical calculations")
def calculate(expression: str):
    """Safely evaluate mathematical expressions."""
    # Your calculation logic here
    return {"result": eval(expression), "expression": expression}

# Sync tools with ElevenLabs (creates/updates as needed)
async def main():
    result = await toolset.sync_tools()
    print(f"Sync result: {result}")
    
    # Start the webhook server
    await toolset.serve(port=8000)

if __name__ == "__main__":
    asyncio.run(main())

Key Features

🎯 Agent-Specific Tools

Tools are automatically prefixed with the agent name (agentName_toolName) and assigned only to the specified agent.

🔄 Intelligent Synchronization

• Compares local tool configurations with remote ones
• Only creates new tools or updates changed ones
• Validates agent existence and name matching

🚀 Zero-Configuration Webhooks

• Automatically generates FastAPI endpoints for each tool
• Handles JSON request/response serialization
• Supports all Python function parameter types

📝 Type-Safe Configuration

• Full TypedDict implementation matching ElevenLabs API schema
• Comprehensive parameter validation
• IDE support with proper type hints

Advanced Usage

Custom Tool Configuration

@toolset.tool(
    name="advanced_search",
    description="Advanced search with custom parameters",
    path="/custom/search",
    response_timeout_secs=30,
    disable_interruptions=True,
    force_pre_tool_speech="enabled",
    auth_connection_id="your_auth_id"
)
def search_database(query: str, limit: int = 10, include_metadata: bool = False):
    """Perform advanced database search."""
    results = perform_search(query, limit, include_metadata)
    return {"results": results, "count": len(results)}

Function Parameter Handling

ElevenTools automatically converts Python function signatures to ElevenLabs tool schemas:

@toolset.tool()
def process_order(
    customer_id: str,           # Required parameter
    items: list,                # Required parameter  
    discount_code: str = None,  # Optional parameter
    priority: bool = False      # Optional parameter with default
):
    """Process customer order with items and optional discount."""
    # Function logic here
    return {"order_id": "12345", "status": "processing"}

Result: Creates a tool named {agentName}_process_order with proper JSON schema validation.

Sync Results

The sync_tools() method returns detailed information about the sync operation:

result = await toolset.sync_tools()
print(result)
# {
#     "status": "success",
#     "agent_id": "agent_xxx",
#     "agent_name": "my_agent", 
#     "created": 2,    # Number of new tools created
#     "updated": 1,    # Number of existing tools updated
#     "total_tools": 5 # Total tools in toolset
# }

Tool Naming Convention

Pattern: {agentName}_{toolName}

Examples:

• Agent: "customer_service" → Tool: "get_weather" → Created as: "customer_service_get_weather"
• Agent: "sales_bot" → Tool: "calculate_discount" → Created as: "sales_bot_calculate_discount"

This ensures tools are properly namespaced and avoid conflicts when multiple agents use similar tool names.

Error Handling

try:
    result = await toolset.sync_tools()
    if result["status"] == "error":
        print(f"Sync failed: {result['message']}")
except Exception as e:
    print(f"Unexpected error: {e}")

API Reference

WebhookToolset

Constructor:

• base_url: Your webhook server's public URL
• xi_api_key: ElevenLabs API key
• agent_id: Target agent ID from ElevenLabs
• agent_name: Agent name for validation and tool prefixing

Methods:

• tool(): Decorator to register functions as tools
• sync_tools(): Sync local tools with ElevenLabs API
• get_tool_configs(): Get list of registered tool configurations
• run(): Start webhook server (synchronous)
• serve(): Start webhook server (asynchronous)

Requirements

• Python 3.8+
• FastAPI
• httpx
• uvicorn
• ElevenLabs API account with conversational AI access

License

MIT License - see LICENSE file for details.

Contributing

Contributions welcome! Please read CONTRIBUTING.md for guidelines.

---

Need help? Check out the examples in the /example directory or open an issue on GitHub.