I’m excited to announce the llm-tools-kiwix plugin: now available for llm.datasette.io. This tool allows you to expose offline Kiwix ZIM archives (think Wikipedia, Stack Exchange, DevDocs, etc.) to Large Language Models (LLMs) via the llm command-line tool or Python API.

Why llm-tools-kiwix?

The ecosystem around LLMs is booming, but most reference tools rely on an internet connection. Enter Kiwix: a project distributing ZIM archives of web sites—like Wikipedia, DevDocs, and Stack Exchange—for true offline access.

Now, you can enable your LLM to tap into these huge offline knowledge bases, even when you’re disconnected, for search and retrieval-augmented generation (RAG) workflows.

Features

  • Automatic ZIM file discovery Finds .zim files in your current working directory or in any folder set by the KIWIX_HOME environment variable.

  • Dynamic tool descriptions for LLMs The list of available ZIM files is automatically reflected in the LLM tools interface—making tool selection smarter.

  • Zero network requirement The plugin works entirely offline, leveraging the power of Kiwix archives.

  • Multiple retrieval tools

    • kiwix_search_and_collect: high-level search and content summarization from a ZIM file.
    • kiwix_search: get search results and metadata.
    • kiwix_read: fetch the content of a specific article.

Quickstart

Install the plugin directly from PyPI:

1

llm install llm-tools-kiwix

How does it work?

  1. Acquire ZIM files Download from Kiwix Downloads. Examples:

    • wikipedia_en_all_nopic_2023-10.zim
    • devdocs_en_docker_2025-04.zim
    • askubuntu.com_en_all_2024-10.zim

  2. Place ZIM files for discovery Put .zim files in your working directory or set a KIWIX_HOME for auto-discovery.

  3. List tools and ZIMs See available tools with detected ZIM files:

    1

    llm tools list

    You’ll find descriptions like:

    1

    Available ZIM files for 'zim_file_path' argument: ./askubuntu.com_en_all_2024-10.zim, ./devdocs_en_docker_2025-04.zim...

Example: Search Docker DevDocs, Offline

Let’s imagine you’re debugging Docker (rootless) and want fast, offline help:

1
2
3

llm -m gpt-4o-mini --tool kiwix_search_and_collect \
  "I'm getting a permission error while trying to run Docker in rootless mode. Please search and provide relevant information from the Docker devdocs." \
  --tools-debug
  • Swap gpt-4o-mini for your model of choice.
  • The LLM will pick up the right ZIM (devdocs_en_docker_2025-04.zim) and walk through the available docs—all offline.

Dev & Contribution

  • Clone the repo and create a virtual environment:
    1
2
3
4
5

    git clone https://github.com/mozanunal/llm-tools-kiwix.git
cd llm-tools-kiwix
python -m venv venv
source venv/bin/activate
pip install -e '.[test]'
  • Run the tests:
    1

    python -m pytest
  • Place ZIM files in your project root or under KIWIX_HOME, or mock the Kiwix backend for test development.

llm-tools-kiwix is open-source (Apache 2.0). Read more or contribute on GitHub.

Unplug, empower your LLM, and take Wikipedia anywhere!