llm-tools-kiwix

Expose offline Kiwix ZIM archives (like Wikipedia, Stack Exchange, DevDocs) to Large Language Models (LLMs) via the LLM CLI tool and Python library. This plugin allows LLMs to search and read content from your local ZIM files.

Key Features

• Automatic ZIM File Discovery: Finds .zim files in the current working directory and in the directory specified by the KIWIX_HOME environment variable.
• Dynamic Tool Descriptions: Tool descriptions are automatically updated with the list of discovered ZIM files, guiding the LLM on what's available.
• Offline Content Access: Enables LLMs to access information from ZIM archives without an internet connection.
• Multiple Tools:
  • kiwix_search_and_collect: Searches a ZIM file and returns the content of matching articles. (Most commonly used)
  • kiwix_search: Performs a search and returns metadata and article paths.
  • kiwix_read: Reads the content of a specific article path.

Installation

Install this plugin in the same environment as LLM.

llm install llm-tools-kiwix

Usage

1. Download ZIM Files: Obtain ZIM files from Kiwix Downloads or other sources. For example, you might download:

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

2. Place ZIM Files: Put the downloaded .zim files in:

   • The directory where you'll be running your llm commands or Python scripts (current working directory).
   • The directory specified by the KIWIX_HOME environment variable, if set. The plugin will automatically detect files from these locations.

   Example (based on your provided snippet):

   # In your project directory
   ls *.zim
   # Expected output (if files are in current directory):
   # askubuntu.com_en_all_2024-10.zim  devdocs_en_docker_2025-04.zim  devdocs_en_scala_2025-04.zim
   

3. Discovering Tools and Available ZIMs: You can see the tools provided by this plugin and which ZIM files they've detected by running:

   llm tools list
   

   This will show entries like kiwix_search_and_collect, and its description will include a line similar to: Available ZIM files for 'zim_file_path' argument: ./askubuntu.com_en_all_2024-10.zim, ./devdocs_en_docker_2025-04.zim, ... The tool's full description will also detail the format of its output.

Command-Line Interface (CLI) Example

You can instruct an LLM to use these tools directly from the command line. For example, if you want to find information about fixing a rootless Docker installation error using your devdocs_en_docker_2025-04.zim file:

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

• Replace gpt-4o-mini with your desired model.
• The LLM will use the prompt to identify that it needs to call the kiwix_search_and_collect tool.
• It will extract zim_file_path: "devdocs_en_docker_2025-04.zim" (or another relevant ZIM file from the available list) and search_string: "docker rootless permission error" (or similar) as arguments for the tool.
• --tools-debug (optional) shows the tool calls and responses.

The LLM will then receive the search results (content from the ZIM file) and use that to formulate its answer.

Development

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

cd llm-tools-kiwix
python -m venv venv
source venv/bin/activate
# On Windows: venv\Scripts\activate

Now install the dependencies and test dependencies:

pip install -e '.[test]'

To run the tests:

python -m pytest

Make sure you have some .zim files in the project root, or in a directory specified by KIWIX_HOME, or mock the ZIM interactions appropriately for tests to run against.