Combining Tools
NPi tools can be combined to create more complex tools.
Combining Tools in Function Mode
When adding a sub-tool to a tool in function mode, the sub-tool's functions are automatically unpacked and added to the parent tool. This allows the parent tool to call the sub-tool's functions as if they were its own, and the parent tool is responsible for the invocation of all functions. (If you wrap the parent tool into an agent, the agent will handle all the reasoning and planning processes.)
The Python code snippet below demonstrates how to combine two tools in function mode. In this case, the sub-tool's add function is directly added to the parent tool and is visible as my_sub_tool__add.
from openai import OpenAI
from npiai import FunctionTool, function
class MySubTool(FunctionTool):
def __init__(self):
super().__init__(
name='my_sub_tool',
description='test sub tool',
)
@function
def add(self, a: int, b: int) -> int:
"""
Add two numbers.
Args:
a: The first number.
b: The second number.
"""
return a + b
class MyTool(FunctionTool):
def __init__(self):
super().__init__(
name='my_tool',
description='test tool',
)
self.add_tool(MySubTool())
@function
def multiply(self, a: int, b: int) -> int:
"""
Multiply two numbers.
Args:
a: The first number.
b: The second number.
"""
return a * b
async def main():
async with MyTool() as my_tool:
print(my_tool.unpack_functions()) # [multiply(), my_sub_tool__add()]Combining Tools in Agent Mode
When combining tools in agent mode, the sub-tool's functions are not automatically unpacked and added to the parent tool. Instead, the sub-tool is treated as a separate agent that can be called by the parent tool via the agent.chat() method, i.e., the sub-tool's agent will be resposible for the reasoning and planning processes for its own functions.
The Python code snippet below demonstrates how to combine two tools in agent mode. In this case, the sub-tool is wrapped as an agent and only the agent's chat function is visible to the parent tool.
from npiai import FunctionTool, agent, function
class MySubTool(FunctionTool):
def __init__(self):
super().__init__(
name='my_sub_tool',
description='test sub tool',
)
@function
def add(self, a: int, b: int) -> int:
"""
Add two numbers.
Args:
a: The first number.
b: The second number.
"""
return a + b
class MyTool(FunctionTool):
def __init__(self):
super().__init__(
name='my_tool',
description='test tool',
)
self.add_tool(
agent.wrap(MySubTool())
)
@function
def multiply(self, a: int, b: int) -> int:
"""
Multiply two numbers.
Args:
a: The first number.
b: The second number.
"""
return a * b
async def main():
async with MyTool() as my_tool:
print(agent.unpack_functions()) # [multiply(), my_sub_tool__agent__chat()]