LMPrep

A lightning-fast utility for preparing and organizing your code for use with LLMs like Claude Projects. LMPrep will collect and rename all of your project files to a flat directory, but preserving the structure within the filenames.

For example, a file at src/models/user.py will be renamed to src^models^user.py in the output directory. Be sure to tell the LLM that your files are structured this way!

demo.mp4

Features

• Smart File Organization: Automatically flattens complex directory structures while preserving path information in the filenames and in a file tree
• Configurable Filtering: Specify which file extensions to include in your dataset to limit context size
• Path Preservation: Uses customizable delimiters to maintain original path information in filenames
• Git-Aware: Respects .gitignore patterns to exclude unwanted files or secrets
• Flexible Output: Generate individual files or create a zip archive
• Visual Tree View: Visualize your source and output file structure, or send the file tree to the LLM
• Fast & Efficient: Written in Rust for maximum performance

Quick Start

Installation

The easiest way to install LMPrep is to get it from PyPi:

pip install lmprep

wheels are built for Windows, Linux, and MacOS.

Manaul / Install Script

1. Download the latest release for your platform from Releases:

   • Windows: lm-x86_64-pc-windows-msvc.zip
   • Linux: lm-x86_64-unknown-linux-gnu.tar.gz
   • macOS: lm-x86_64-apple-darwin.tar.gz

2. Install the binary:

Linux/macOS:

# Extract and copy binary
tar xzf lm-x86_64-*-*.tar.gz
sudo mv lm /usr/local/bin/

# Create config file
curl -O https://raw.githubusercontent.com/bcherb2/lmprep/main/src/config-example.yaml
mv config-example.yaml ~/.lmprep.yml

Windows (in PowerShell, run as Administrator):

# Extract and copy binary
Expand-Archive lm-x86_64-pc-windows-msvc.zip
New-Item -ItemType Directory -Force -Path "C:\Program Files\lmprep"
Move-Item -Force lm.exe "C:\Program Files\lmprep\lm.exe"
$env:Path += ";C:\Program Files\lmprep"

# Create config file
Invoke-WebRequest -Uri "https://raw.githubusercontent.com/bcherb2/lmprep/main/src/config-example.yaml" -OutFile "$env:USERPROFILE\.lmprep.yml"

3. Verify installation:

lm --help

Alternative: Build from Source

If you have Rust installed, you can build from source:

git clone https://github.com/bcherb2/lmprep
cd lmprep
cargo build --release

The binary will be in target/release/lm (or lm.exe on Windows). Follow step 2 above to set up the config file.

Basic Usage

# Create a default config file in the current directory
lm --init-config

# Organize files in current directory
lm .

# Organize files from a specific directory
lm /path/to/source

# Use a custom config file
lm . -c /path/to/.lmprep.yml

# Create a zip archive instead of of individual files
lm . --zip

Configuration

Create a .lmprep.yml file in your home directory to customize behavior, or create one in your project root directory. Here's an example:

allowed_extensions:
  - py
  - rs
  - md
  - txt
ignored_directories:
  - node_modules
delimiter: "^"
subfolder: context
zip: false
tree: true
respect_gitignore: true

NOTE: The install script will create a default config file at ~/.lmprep.yml

Configuration Options

Option	Description	Default
allowed_extensions	File extensions to include	[] (common extensions)
ignored_directories	Directories to ignore	[] (common directories)
delimiter	Character used to represent path hierarchy	^
subfolder	Output directory name within project	context
zip	Create zip archive instead of files	false
tree	Show file tree visualization	true
respect_gitignore	Honor .gitignore patterns	true

Command Line Options

lm [OPTIONS] [SOURCE]

Arguments:
  [SOURCE]  Source directory to organize files from [default: .]

Options:
  -c, --config <FILE>     Path to config file
  -s, --subfolder <NAME>  Override the subfolder name from config
  -z, --zip              Create a zip file instead of individual files
  -t, --tree             Show file tree of source and output
  -v, --verbose          Show more detailed output during processing
      --init-config      Create a default config file in the current directory
  -h, --help             Print help
  -V, --version          Print version

Development

To set up for development:

1. Clone the repository
2. Run ./dev-setup.sh

This will build the Rust binary, set up the correct directory structure, and install the package in development mode.

When you make changes to the Rust code, run ./dev-setup.sh again to rebuild and reinstall. Python changes will be picked up automatically due to the development install.

Use Cases

• Code Analysis: Organize your code into a flat structure while preserving context (works especially well with Claude Projects)
• Document Processing: Organize and prepare document collections for processing, logs, etc.
• Version Control: Easily create clean snapshots of your codebase for archival in zip format

Building from Source

1. Install Rust using rustup
2. Clone the repository:

   git clone https://github.com/bcherb2/lmprep.git
   cd lmprep

3. Build the project:

   cargo build --release

4. The binary will be available at target/release/lm, copy it and add it to your PATH
5. Create the .lmprep.yml file in your home directory or project root

NOTE: see install/BUILD.md for more in depth building instructions.

FAQ

Q: Why use LMPrep instead of just copying files? A: LMPrep preserves directory structure information in filenames, making it easier for LLMs to understand file relationships and context. Sure, you can do this manually, but it gets tedious.

Q: How does path flattening work? A: A file at src/models/user.py becomes src^models^user.py in the output directory (using default delimiter). Changing the delimiter to + would result in src+models+user.py.

Q: Can I exclude certain files or directories? A: Yes! LMPrep respects .gitignore patterns and allows you to specify allowed file extensions.

Q: Is it safe to use on large directories? A: Yes! LMPrep is written in Rust for performance and memory efficiency, making it suitable for large datasets.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.