Enhance Your RAG Application With Web Searching Capability!

Introduction

When building fun projects with Retrieval-Augmented Generation (RAG) applications, we often face limitations like browsing restrictions, making it hard to get the latest information or current data, like weather updates (i hope something more funny). To solve this, we can equip our RAG application with tools to search the internet. Let’s dive in!

 Our Tool-bench

• LangChain (Framework for building applications with large language models)
• SearXNG (free metasearch engine)
• CPython (a C language wrapper :> )
• Docker

Setup

First we start with the SearXNG installation.

1 -) Get SearXNG-docker

git clone https://github.com/searxng/searxng-docker.git

2 -) Edit the .env file to set the hostname and an email

3 -) Generate the secret key

<Linux>

sed -i "s|ultrasecretkey|$(openssl rand -hex 32)|g" searxng/settings.yml

<MacOS>
sed -i"" -e "s|ultrasecretkey|$(openssl rand -hex 32)|g" searxng/settings.yml 

<Windows>
$randomBytes = New-Object byte[] 32
(New-Object Security.Cryptography.RNGCryptoServiceProvider).GetBytes($randomBytes)
$secretKey = -join ($randomBytes | ForEach-Object { "{0:x2}" -f $_ })
(Get-Content searxng/settings.yml) -replace 'ultrasecretkey', $secretKey | Set-Content searxng/settings.yml

4 -) Update the searxng/settings.yml to enable available search formats and disable the limiter for our LangChain instance:

use_default_settings: true
server:
  # base_url is defined in the SEARXNG_BASE_URL environment variable, see .env and docker-compose.yml
  secret_key: "<secret-key>"  # change this!
  limiter: false
  image_proxy: true
ui:
  static_use_hash: true
redis:
  url: redis://redis:6379/0

search:
    formats:
        - html
        - json

5-) Run SearXNG Instance

docker compose up

Check the SearXNG deployment in Docker. If everything looks good, you’re ready to continue.

 Demo Application

1 -) Create a virtual environment & activate

python3 -m venv .venv
source .venv/bin/activate

2 -) Install Langchain

pip install langchain langchain-community

3 -) Create main.py

## Simple Get Results
from langchain_community.utilities import SearxSearchWrapper
import pprint

s = SearxSearchWrapper(searx_host="http://localhost:8080",)
result = s.results("What is RAG?", num_results=10, engines=["google"])
pprint.pprint(result)

## Github Tool

from langchain_community.tools.searx_search.tool import SearxSearchResults

wrapper = SearxSearchWrapper(searx_host="**")
github_tool = SearxSearchResults(name="Github", wrapper=wrapper,
                            kwargs = {
                                "engines": ["github"],
                                })

And there you have it! Your RAG application now has search capabilities. This guide doesn’t introduce anything new but aims to bring together the steps for adding web searching functionality to your RAG application. I hope it helps!