feat: add tool support for the agent

[DagaBhai]

What does this PR do?

contributing to #104

adds a tool support for agent

files change:
tools.py - tool decorator for registering tool with the agent
node.py/ActNode - added actual executable tool calling as well as if any error happens agent will get to know
agent.py - shared dict added "tools": get_tool_schema() for the agent to know more about function and its input

Test

from fenn.agents.tools import tool, get_tool_schema, execute_tool
from fenn.agents.node import ActNode

@tool
def calculate_calculator(expression: str) -> str:
    """
    Evaluates a mathematical expression string safely.
    Use this when you need to do basic math like addition, multiplication, or division.
    """
    try:
        # Note: In production, use a safer eval alternative like simpleeval
        return str(eval(expression, {"__builtins__": None}, {}))
    except Exception as e:
        return f"Error evaluating expression: {str(e)}"

@tool
def fetch_weather(location: str, date:str) -> str:
    """
    Retrieves the current weather for a specific city location.
    """
    # Simulate a real API call or database lookup
    return f"The weather in {location} on {date} is sunny and 22°C."


print("Available tools for the LLM:")
print(get_tool_schema())

# 2. When the agent outputs a tool call, execute it
tool_call = "fetch_weather(xyz, 20-05-2026)"
thought = """
Think: I should check weather
Action: fetch_weather(xyz, 20-05-2026)
"""

result1 = ActNode.exec(ActNode(),thought=tool_call)
result2 = ActNode.exec(ActNode(),thought=thought)
print(f"\nExecution Result1: {result1}")
print(f"\nExecution Result2: {result2}")

Note

• tool calling can only be enforced if agent has correct observation of the envirnoment, output/result will let it know if the error has happened or not to do that llm must give json or any structured output
• If an error occurs, the error message should be passed back to the agent, as the LLM tends to provide a better fix once it knows what went wrong.

[blkdmr]

Thank you for your help!