Linux Command LLM

Overview

Linux Command LLM is a fine-tuned language model designed to interpret natural language commands and generate corresponding Linux shell commands. This model is based on open_llama_3b_v2 and has been fine-tuned to assist users in efficiently executing system operations through natural language input.

Features

• Natural Language to Shell Command Translation: Converts human-readable instructions into precise Linux commands.
• Fine-Tuned for Linux Tasks: Covers command-line utilities, file management, networking, package management, and system monitoring.
• Optimized for WSL ARM: Ensuring seamless execution on ARM-based Windows Subsystem for Linux (WSL).
• Integration with LLM Kernel: Can be embedded into an OS with an LLM-integrated kernel for intuitive system interactions.

Model Information

• Base Model: open_llama_3b_v2
• Fine-Tuned Model: Hugging Face Repository
• Parameter Size: 3B

Setup & Installation

Prerequisites

• Python 3.8+
• PyTorch
• Hugging Face Transformers Library
• WSL ARM Environment (for deployment on WSL ARM)

Installation Steps

1. Clone the Repository

   git clone https://github.com/your-repo/linux_command_llm.git
   cd linux_command_llm

2. Create a Virtual Environment

   python -m venv venv
   source venv/bin/activate  # On Windows, use venv\Scripts\activate

3. Install Dependencies

   pip install -r requirements.txt

4. Download the Fine-Tuned Model

   from transformers import AutoModelForCausalLM, AutoTokenizer
   
   model_name = "kushagragoyal/fine_tuned_model"
   tokenizer = AutoTokenizer.from_pretrained(model_name)
   model = AutoModelForCausalLM.from_pretrained(model_name)

Usage

Running the Model

from transformers import pipeline

pipe = pipeline("text2text-generation", model=model, tokenizer=tokenizer)
command = pipe("List all files in the current directory")
print("Generated Command:", command[0]['generated_text'])

Example Queries

Input (Natural Language)	Output (Shell Command)
"List all files in a directory"	ls -l
"Show the current working directory"	pwd
"Check disk usage"	df -h
"Find a file named 'log.txt'"	find / -name 'log.txt'

Roadmap

• Phase 1: Basic Linux command support (✓ Completed)
• Phase 2: Context-aware command generation
• Phase 3: Support for multi-command workflows
• Phase 4: Full OS integration with LLM kernel

Contribution

Contributions are welcome! If you’d like to improve the model, fine-tune additional commands, or integrate new features, feel free to submit a pull request.

License

This project is licensed under the MIT License. See LICENSE for details.