all4llm

A command-line tool that assembles multiple source code files (by recursively scanning directories among other ways) into a single string, in a format suitable to send to Large Language Models (LLMs). It automatically:

• Combines multiple source files into a single string with clear demarcation between the sections representing each file.
• Supports direct clipboard copying for easy pasting into LLM chats
• Recursively scans directories to gather the files
• Handles common source code files by default
• Basic filtering for including / excluding files by extension
• Estimates token count for LLM context planning

Installation

# Clone the repository
git clone https://github.com/codr1/all4llm.git
cd all4llm

# Make the script executable
chmod +x all4llm

# Optional: Add to your PATH
sudo ln -s $(pwd)/all4llm /usr/local/bin/all4llm

# For clipboard support on Linux:
sudo apt-get install xclip  # Ubuntu/Debian
sudo dnf install xclip      # Fedora
sudo pacman -S xclip        # Arch
# Note: macOS clipboard support works out of the box

Usage

Basic syntax:

./all4llm [OPTIONS] DIRECTORY

Common Use Cases

# View source files in current directory
./all4llm .

# Copy all source files to clipboard (for pasting into ChatGPT, etc.)
./all4llm . -c

# Process subdirectories recursively
./all4llm -r src/

# Save output to a file
./all4llm -r . -o project_code.txt

# Only include Python files
./all4llm . --only "*.py"

# Include multiple specific files/patterns
./all4llm . --only "file1.py" --only "file2.js" --only "config.json"

# Add custom file patterns
./all4llm . -i "*.conf" -i "*.cfg"

Options

• -h, --help: Show help message
• -r, --recursive: Include subdirectories recursively
• -i, --include PATTERN: Include additional file pattern
• -o, --output FILE: Write output to file
• -c, --clipboard: Copy output to clipboard
• --no-hidden: Skip hidden files and directories
• --list-defaults: Show default file extensions
• --only PATTERN: Only include specified pattern (can be used multiple times, disables defaults)

Default File Types

Automatically includes common source code files:

• C/C++: .c, .h, .cpp, .hpp, .cc
• Python: .py, .pyw
• JavaScript/TypeScript: .js, .ts, .jsx, .tsx
• Go: .go
• Java: .java
• Rust: .rs
• Ruby: .rb
• PHP: .php
• Documentation: .md, .rst
• Config files: .json, .yaml, .yml
• And more... (use --list-defaults to see all)

Output Format

Files are marked in the output like this:

<<<< FILE: path/to/file.ext >>>>
[file contents]
<<<< END FILE: path/to/file.ext >>>>

The last line of output always shows an estimate of how many tokens the content will consume in your LLM context window.

Tips

• Use -c for direct clipboard copy so you can easily interract with an LLM
• Token count helps you stay within LLM context limits
• Use --only "*.ext" to focus on specific file types
• Use -r when you need to include subdirectories