llm-tools-searxng

A tool to search the web using SearXNG search engines.

Installation

Install this plugin in the same environment as LLM.

llm install llm-tools-searxng

Configuration

By default, the tool does not have a default SearXNG URL set. You can configure it in one of two ways:

Using environment variables

export SEARXNG_URL=https://your-searxng-instance.com
export SEARXNG_METHOD=GET  # or POST (default)

Using LLM's built-in key management

llm keys set searxng_url --value https://your-searxng-instance.com

Note: Public SearXNG instances typically don't allow API access or JSON output.

Usage

Simple search function

Use the searxng_search function for basic web searches:

llm --tool searxng_search "latest developments in AI" --tools-debug

With LLM chat

This plugin works well with llm chat:

llm chat --tool searxng_search --tools-debug

Python API usage

import llm
from llm_tools_searxng import SearXNG, searxng_search

# Using the simple function
model = llm.get_model("gpt-4.1-mini")
result = model.chain(
    "What are the latest developments in renewable energy?",
    tools=[searxng_search]
).text()

Development

To set up this plugin locally, first checkout the code. Then create a new virtual environment:

cd llm-tools-searxng
uv sync --all-extras

Now install the dependencies and test dependencies:

llm install -e '.[test]'

To run the tests:

uv run python -m pytest