Making LLM Web Search Easy: A 5-Minute Guide

Photo by Growtika on Unsplash

Building an LLM that can search the web is surprisingly straightforward. This guide will show you how to do it in just five minutes.

Installation: Quick Setup

First, install the necessary Python packages:

%%capture
%pip install -qU langchain-groq
%pip install -qU langchain
%pip install -qU langchain-community
%pip install -qU duckduckgo-search

LLM Setup: Leveraging Groq’s Speed

We’ll use Groq for this example due to its speed. Here’s the setup code:

import os
from langchain_groq import ChatGroq

os.environ["GROQ_API_KEY"] = "API_KEY"

llm = ChatGroq(
    model="mixtral-8x7b-32768",
    temperature=0,
    max_tokens=None,
    timeout=None,
    max_retries=2,
)

This code initializes the ChatGroq LLM, setting parameters for model, temperature, and retries.

Tool Integration: DuckDuckGo Search

Next, we’ll integrate the DuckDuckGo search engine as a tool:

from langchain.agents import initialize_agent
from langchain_community.tools import Tool, DuckDuckGoSearchRun, DuckDuckGoSearchResults
from langchain.agents import Tool

ddg_search = DuckDuckGoSearchResults()

tools = [
   Tool(
       name="DuckDuckGo Search",
       func=ddg_search.run,
       description="Useful to browse information from the Internet.",
   )
]

agent = initialize_agent(
   tools, llm, agent="zero-shot-react-description", verbose=True
)

This snippet defines the search tool and initializes the agent with the LLM and tools.

Witness the Magic: Running the Agent

Finally, run the agent with a simple query:

from langchain.agents import initialize_agent

agent = initialize_agent(
   tools, model, agent="zero-shot-react-description", verbose=True
)

agent.run(
   "Weather in London in the coming 3 days"
)

This line of code executes the query using the DuckDuckGo search tool and the Groq LLM.

Conclusion

Simple and PowerfulCreating an LLM that can search the web is remarkably easy with these steps. Explore the possibilities and build your own web-searching LLM!

The full code is available below:

hungrydata/llm_with_search.ipynb at main · morriswong/hungrydata