Custom agent logic in LangChain - Mini Project: Build & Apply

Custom Agent Logic with LangChain

📖 Scenario: You are building a simple AI assistant using LangChain. This assistant will decide how to respond based on the user's input by using custom agent logic.

🎯 Goal: Create a LangChain agent with custom logic that chooses between two tools based on the user's question.

📋 What You'll Learn

Create a dictionary called tools with two tools: 'calculator' and 'search'

Create a variable called agent_config with a key use_search set to true

Write a function called custom_agent_logic that takes input_text and returns 'search' if agent_config['use_search'] is true and the input contains the word 'weather', otherwise returns 'calculator'

Create an agent instance called agent that uses custom_agent_logic to select the tool

💡 Why This Matters

🌍 Real World

Custom agents let you build AI assistants that decide how to respond based on the question, like choosing between searching the web or doing calculations.

💼 Career

Understanding custom agent logic is useful for AI developers building smart assistants, chatbots, or automation tools that need flexible decision-making.

1

Create the tools dictionary

Create a dictionary called tools with two entries: 'calculator' mapped to CalculatorTool() and 'search' mapped to SearchTool().

LangChain

# Your code here

Need a hint?

Use a Python dictionary with keys 'calculator' and 'search' and assign the respective tool instances.

2

Add agent configuration

Create a dictionary called agent_config with a key 'use_search' set to true.

LangChain

tools = {
    'calculator': CalculatorTool(),
    'search': SearchTool()
}
# Your code here

Need a hint?

Define a dictionary named agent_config with the key 'use_search' set to true.

3

Write the custom agent logic function

Write a function called custom_agent_logic that takes a parameter input_text. The function should return 'search' if agent_config['use_search'] is true and the string 'weather' is in input_text (case insensitive). Otherwise, return 'calculator'.

LangChain

tools = {
    'calculator': CalculatorTool(),
    'search': SearchTool()
}

agent_config = {'use_search': True}

# Your code here

Need a hint?

Use input_text.lower() to check for 'weather' ignoring case. Use an if-else statement to return the correct tool name.

4

Create the agent with custom logic

Create a variable called agent and assign it an instance of CustomAgent initialized with tools and custom_agent_logic as the logic function.

LangChain

tools = {
    'calculator': CalculatorTool(),
    'search': SearchTool()
}

agent_config = {'use_search': True}

def custom_agent_logic(input_text):
    if agent_config['use_search'] and 'weather' in input_text.lower():
        return 'search'
    else:
        return 'calculator'

# Your code here

Need a hint?

Instantiate CustomAgent with the tools dictionary and the custom_agent_logic function.