codex-apply-patch

Codex Apply Patch

A CLI tool and Python library for applying patches using a custom patch format designed for AI coding assistants.

This project is based on OpenAI's original codex apply-patch tool and extends it with additional functionality.

Changes from Original

This fork adds the following enhancements to the original OpenAI tool:

• In-Memory Patch Application: New functionality to apply patches to files in memory without touching the filesystem, useful for testing and preview scenarios.

• Python Library Support: Complete Python bindings using PyO3, allowing the patch functionality to be used directly from Python code.

• PyPI Distribution: The library is packaged and distributed via PyPI for easy installation.

Installation

Python Package

pip install codex-apply-patch

From Source

# Install Rust if you haven't already
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

# Clone and build
git clone https://github.com/openai/codex-apply-patch
cd codex-apply-patch
cargo build --release

# For Python development
pip install maturin
maturin develop --release

Usage

Command Line

# Apply patch from stdin
echo "*** Begin Patch
*** Add File: hello.txt
+Hello, World!
*** End Patch" | codex_apply_patch

Python Library

import codex_apply_patch as cap

# Apply patch to files on disk
patch = """*** Begin Patch
*** Add File: hello.py
+print("Hello, World!")
*** End Patch"""

result = cap.apply_patch(patch)
print(result)

# Apply patch in memory (new feature)
files = {
    "main.py": "def main():\n    print('old version')\n"
}

patch = """*** Begin Patch
*** Update File: main.py
@@
 def main():
-    print('old version')
+    print('new version')
*** End Patch"""

result = cap.apply_patch_in_memory(patch, files)
print("Modified files:", result.files)
print("Summary:", f"Added: {len(result.added)}, Modified: {len(result.modified)}, Deleted: {len(result.deleted)}")

# Generate patch from file contents (new feature)
original = "def hello():\n    print('old version')"
new = "def hello():\n    print('new version')\n    print('updated!')"
patch = cap.generate_patch("main.py", original, new)
print("Generated patch:")
print(patch)

# Generate patch for multiple files (new feature)
file_changes = {
    "new.py": (None, "print('new file')"),  # Add file
    "old.py": ("old content", None),        # Delete file
    "update.py": ("old", "new")             # Update file
}
multi_patch = cap.generate_patch_from_files(file_changes)
print("Multi-file patch:")
print(multi_patch)

Python API Reference

• apply_patch(patch_str) - Apply patch to files on disk
• apply_patch_in_memory(patch_str, files_dict) - Apply patch to in-memory files
• parse_patch(patch_str) - Parse patch and return structure information
• generate_patch(path, original_content, new_content) - Generate patch for a single file
• generate_patch_from_files(files_dict) - Generate patch for multiple files
• get_tool_instructions() - Get the CLI tool instructions for AI assistants
• get_api_instructions() - Get the patch format instructions (without CLI specifics)

License

Licensed under the Apache License, Version 2.0. See LICENSE for details.

This project extends the original OpenAI codex apply-patch tool, which is also licensed under Apache-2.0.