Tools in LangChain

What tools are, how a tool call flows, the tool types and popular built-ins, and how to create custom tools and use several with an agent.

07Agents in LangChain 10Agents in LangChain (Advanced)17LangChain Chains, Agents and Tools

9.1

What are Tools?

Tools allow LLMs to interact with external systems (APIs, databases, search engines, etc.).
LLMs can call tools to get real-time information, perform actions or execute code.
LangChain provides a standard interface to create, integrate and manage tools.
Tools are essential for building agents and real-world applications.

9.2

Why Use Tools?

Access real-time and external information.
Perform actions like search, calculation and API calls.
Extend LLM capabilities beyond training data.
Useful for building RAG, agents and automation.
Easily create custom tools.
Integrate with popular services (Google Search, Wikipedia, Python REPL, databases, etc.).

9.3

How Tools Work in LangChain

User Query

LLMDecides to use a tool

ToolExternal system / API / DB

Tool ResultReturns data to LLM

Final ResponseTo user

9.4

Types of Tools in LangChain

Tool Type	Description	Use Case
Built-in Tools	Pre-built tools provided by LangChain	Search, Wikipedia, Python REPL, etc.
API Tools	Call external APIs	Weather, finance, custom APIs
Database Tools	Interact with databases	Query MySQL, PostgreSQL, etc.
Custom Tools	Create your own tool using a function	Business logic, internal services
Toolkits	Collection of related tools	SerpAPI toolkit, SQL toolkit, etc.

9.5

Popular Built-in Tools

SerpAPIWrapper - Google Search
WikipediaQueryRun - Wikipedia search
PythonREPLTool - Execute Python code
RequestsToolkit - Make HTTP requests
SQLDatabaseToolkit - Work with SQL databases
ArxivQueryRun - Search research papers
DuckDuckGoSearchRun - Web search (open source)
OpenWeatherMap - Weather information
... and many more via the LangChain community

9.6

Creating a Custom Tool

Define a function with input and output.
Use the @tool decorator or BaseTool.
Provide a clear description for the LLM.
It can be used with agents or chains.

9.7

Example 1: SerpAPI Google Search Tool

python

from langchain.agents import initialize_agent, AgentType
from langchain_community.utilities import SerpAPIWrapper
from langchain.tools import Tool
from langchain_openai import ChatOpenAI

# Initialize LLM
llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.7)

# Create search tool
search = SerpAPIWrapper()
tools = [Tool(
    name="Google Search",
    func=search.run,
    description="Search the internet for latest information")
]
# Initialize agent
agent = initialize_agent(tools, llm,
    agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
    verbose=True)

response = agent.invoke({"input": "Who won the IPL 2024?"})
print(response["output"])

9.8

Example 2: Custom Tool (Calculator)

python

from langchain.tools import tool
from langchain.agents import initialize_agent, AgentType
from langchain_openai import ChatOpenAI

@tool
def calculator(expression: str) -> str:
    """Useful for mathematical calculations.
    Input should be a valid Python expression."""
    try:
        return str(eval(expression))
    except Exception as e:
        return f"Error: {str(e)}"

llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.7)
tools = [calculator]

agent = initialize_agent(tools, llm,
    agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
    verbose=True)

response = agent.invoke({"input": "What is 25 * 4 + 10?"})
print(response["output"])

9.9

Example 3: Using Multiple Tools with an Agent

python

from langchain_community.utilities import SerpAPIWrapper
from langchain.tools import Tool

search_tool = Tool(name="Search", func=SerpAPIWrapper().run,
    description="Search the internet")
calc_tool = calculator  # from previous example

tools = [search_tool, calc_tool]
agent = initialize_agent(tools, llm,
    agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
    verbose=True)

9.10

Key Takeaways

Tools extend LLM capabilities to real-world applications.
Use built-in tools for common tasks.
Create custom tools for business-specific needs.
Tools are the building blocks for powerful agents.
Provide clear descriptions for better tool selection.
Combine multiple tools for complex workflows.
Experiment with different tools and use cases.