gpt4all-code-review 0.13

A standalone code review tool based on GPT4ALL.

Code Review Automation Tool

This program is designed to assist developers by automating the process of code review. By leveraging a pre-trained standalone machine learning model (e.g., GPT-4), it reads source code files and provides suggestions for improvements.

Features

• File Scanning: Ability to scan a single file or all files in the current directory.
• Multiple Output Formats: Results can be exported as plain text, JSON, or XML files.
• Customizable Model: Utilizes a customizable language model, allowing users to select the appropriate model for their needs.

Installation

Make sure you have Python 3.x installed.

pip install -r requirements.txt

or

pip3 install -r requirements.txt

Usage

You can run the program from the command line with various options:

gpt4all_code_review --file <file_path> --all

Options

• h, --help: Show this help message and exit
• --model: Specifies the model name. Default is "orca-mini-3b.ggmlv3.q4_0.bin".
• --file: Specifies the file path to analyze. If not provided, all files in the current directory will be analyzed.
• --all: Includes all files and folders in the current directory for scanning.
• --output: Output type (default: txt). Options: txt, json, xml.
• --export: Export to file (default: False).
• --export-folder: Export to folder (default: ./code_review_results).

Example

To analyze a single file:

gpt4all_code_review --file=./path/to/yourfile.py

To analyze all files in the current directory:

gpt4all_code_review --all

Output Formats

• Text: A human-readable table with file paths and suggestions.
• JSON: A machine-readable format that can be parsed programmatically.
• XML: An alternative machine-readable format.

Dependencies

• gpt4all: Library used for utilizing the GPT-4 model.
• prettytable: Library used for rendering results in a table format.

License

MIT3