GitHub Code Search CLI (ghcs) - With Semantic Code Refinement

ghcs is a powerful command-line interface (CLI) tool for searching code on GitHub and downloading matched files. It allows you to search for code snippets using various filters such as language, user, repository, and path. Additionally, ghcs enables AI-powered code manipulation and refinement using large language models (LLMs) like Gemini. This makes it especially useful for developers working in CLI-based environments without a graphical user interface (e.g., remote servers).

Features

• Search GitHub repositories for code using advanced filters (query, user, repository, language, path, etc.).
• Download matched files directly from GitHub.
• Extract, modify, enhance, and refine code using AI-powered transformations with models like Gemini.
• User-friendly CLI interface designed for seamless integration into developer workflows.

Installation

To install ghcs, use pip:

pip install ghcs

For the latest updates, install directly from GitHub:

pip install git+https://github.com/hissain/ghcs.git

Usage

To use ghcs, you need a GitHub Personal Access Token, which can be set using the --token argument or the GITHUB_TOKEN environment variable. For AI-powered code manipulation with --remark, you must set the GEMINI_API_KEY in your environment.

Basic Search

ghcs 'search_term' --token YOUR_GITHUB_TOKEN

ghcs 'search_term' # When GITHUB_TOKEN is already set in .env

Search with Filters

ghcs 'search_term' --user 'username' --repo 'username/repo' --language 'python' --path 'llama/train' --token YOUR_GITHUB_TOKEN --max-results MAX_RESULT_COUNT

Download Matched Files

ghcs 'search_term' --download --token YOUR_GITHUB_TOKEN

AI-Powered Code Extraction & Refinement

To extract specific code sections or apply AI-driven transformations on downloaded files:

• Use --remark to specify semantic modifications (requires --download).
• The refined code can be printed to the console or saved using --output-file.

Example:

ghcs "LoRA def train()" --user hissain --download --remark "Extract the training function for LoRA with proper imports" --output-file extracted_code.py --verbose

Note: The GEMINI_API_KEY must be set in your environment variables or .env file to enable the --remark feature.

CLI Arguments

Positional Arguments

• query: Search term as a string (required).

Optional Arguments

• -l, --language : Filter by programming language.
• -u, --user : Search within all repositories of a specific user.
• -r, --repo : Search within a specific repository (e.g., username/repo).
• -p, --path : Restrict search to a specific file path.
• -t, --token : GitHub Personal Access Token (or set GITHUB_TOKEN environment variable).
• -m, --max-result : Limit the number of search results.
• -d, --download : Download matched files.
• -dd, --download-dir : Specify the directory for downloaded files.
• -v, --verbose : Enable verbose logging.
• -r, --remark : AI instruction for refining downloaded files.
• -o, --output-file : Output file to save refined code (default: print to console).
• -e, --extensions : Specify file extensions to consider (e.g., .py,.js).
• -h, --help : Show help menu and exit.

Example Commands

ghcs 'def train(' --language 'python' --user 'hissain' --path 'llama/train' --download --token YOUR_GITHUB_TOKEN --max-results 5

ghcs 'def train(' -l 'python' -p 'llama/train' -d -m 10

ghcs "def train()" --path llm --max-results 3 --download

With AI-powered refinement:

ghcs "def train LoRA" --path llm --download --remark "Extract only the forward pass function" --output-file forward_pass.py --max-results 5

API Keys

• GitHub Token: Generate a personal access token at GitHub Tokens
• Gemini API Key: Obtain from Google AI Studio

License

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

Author

Md. Sazzad Hissain Khan

• Email: hissain.khan@gmail.com
• GitHub: hissain

Feel free to modify and enhance this project as needed!