llmcatz 🐱

A cute and powerful tool to gather files into a format perfect for LLMs and AI assistants.

              |\      _,,,---,,_
        ZZZzz /,`.-'`'    -.  ;-;;,_
             |,4-  ) )-,_. ,\ (  `'-'
            '---''(_/--'  `-'\_) 

What is llmcatz?

llmcatz is a lightning-fast utility that scans your codebase and creates a structured output optimized for AI assistants like Claude, ChatGPT, or other LLMs. It formats your file structure and contents, optionally counts tokens, and supports clipboard integration for easy pasting.

Features

• 🚀 Super Fast: Written in Zig for maximum performance.
• 🧵 Multi-threaded: Processes files in parallel (customizable thread count).
• 📋 Clipboard Integration: Copies results to your clipboard (X11/Wayland).
• 🔍 Flexible Targeting: Process files, directories, or GitHub repositories.
• 🙈 Exclusion Patterns: Skip files or directories you don’t want.
• 🔎 Interactive Selection: Use fzf to pick files interactively.
• 🧮 Token Counting: Count tokens using TikToken encodings (e.g., cl100k_base).
• 📊 JSON Export: Export results in JSON format for programmatic use.
• 🐱 Adorable ASCII Art: Because why not?

Dependencies

• Zig (for building the main binary)
• Rust (for building the tiktoken_ffi library)
• fzf (optional, for interactive file selection)
• xclip or wl-copy (optional, for clipboard support on X11 or Wayland)

Installation

1. Install Dependencies:

   • Zig: Follow official instructions.
   • Rust: Install via rustup.
   • fzf, xclip, or wl-copy: Use your package manager (e.g., apt, brew).

2. Clone and Build:

   git clone https://github.com/0xdilo/llmcatz
   cd llmcatz
   ./build.sh
   

3. Optionally, install globally:

    sudo mv zig-out/bin/llmcatz /usr/local/bin/
   

Usage

llmcatz [OPTIONS] [TARGETS...]

# Interactive file selection with fzf
llmcatz -f
# Print to stdout
llmcatz -p src/
# Save to file
llmcatz -o output.txt src/
# Exclude patterns
llmcatz -e ".git,node_modules" src/
# Set thread count
llmcatz -t 8 src/
# Process specific files
llmcatz file1.txt file2.txt
# Clone and process a GitHub repository
llmcatz https://github.com/username/repo
# Use a specific token encoding
llmcatz --encoding o200k_base src/
# Count total files processed
llmcatz --count-files src/
# Count tokens only
llmcatz --count-tokens src/
# Export to JSON format
llmcatz --json -o output.json src/