Club

CLI wrapper for the caveman_compression by wilpel. The name of the tool (club) is a kinda of a joke since this wrapper is a CLI tool and the club was one of the caveman's main tools.

This project was born simply from my curiosity in trying to develop a tool for the CLI. I made some small updates to the original scripts (see updates sections) but all credit for the original code goes to original authors.

List of Contents

• Installation
• Usage
• Examples
• Updates

Installation

To install club you will first need pipx (you can easily install it through brew, pip or scoop based on your system/preferences)

brew install pipx

pip install pipx

scoop install pipx

pipx ensurepath

After installing pipx you can simply install club using the command

pipx install caveman-club

(Don't ask me why but when I upload the packet on Pypi the name club was not allowed for a package, although I couldn't find anyone else using it)

Usage

Club offers the same functionality as the base package but wrapped in a command for CLI. It basically allows you to compress/decompress text through 3 techniques: NLP, MLM and LLM (see here for details about the three methods).

The structure of a command is the following :

club TYPE MODALITY [TEXT] OPTIONS

Where :

• TYPE is on the 3 techniques implemented (llm, mlm and nlp).
• MODALITY can be compress or decompress
• [TEXT] is the text you want to compress or decompress.
• OPTIONS are optional flags. Some flags are common between all 3 TYPE while others are specific to some of them (see flags sections below for more details).

E.g. if you want to compress your text with nlp then you can run

club nlp compress "your verbose text"

E.g. if you want to decompress your text with nlp then you can run

club nlp decompress "your verbose text"

For more examples see the dedicated section below.

N.B. 1. If you do not specify TYPE and write directly the MODALITY, then club will default to nlp. I.e. running club compress "your verbose text" is equivalent to running club nlp compress "your verbose text".

N.B. 2. [TEXT] is an optional parameter you if you specify input and output files. See common flags and examples sections below.

Common Flags

• --input : The input file with the text you want to compress/decompress. If passed together with text, the text will take precedence and a copy of the text will be saved to the input file (backup). If a file already exists at the input file path, it will be overwritten without warning.
• --output : The output file where you want to save the output
• --verbose : Enable verbose output (TYPE with eventual settings, input, and output, and compression ratio if MODALITY is compression)

nlp Flags

• --language : Language code. If not specified it will default to en. Possible values are [en|es|de|fr|it|pt|nl|el|nb|lt|ja|zh|pl|ro|ru].

mlp Flags

Examples

Updates

This is a list of small changes/updates from the original scripts

• Add support for metal (MacOs) for the mlm method
• Changed spacy models import to make it automated the first time the tool is used

TODO

• Add support for other LLMs providers.