CodeQUEST (Code Quality Understanding and Enhancement System Toolkit)
CodeQUEST is a novel framework utilizing Large Language Models (LLMs) to assess and improve code across various quality dimensions, such as readability, security, maintainability, and efficiency. Inspired by the Actor-Critic method from Reinforcement Learning, our framework has two components:
- An Evaluator (Critic) that examines ten dimensions of code quality to provide both quantitative and qualitative evaluations.
- An Optimizer (Actor) that improves the code based on these evaluations. In the code improvement cycle, the Optimizer grounds its improvement on the Evaluator's feedback, seeking higher code quality across multiple code quality dimensions jointly.
Ensure you have Python version 3.10 or higher installed on your system.
git clone <url>python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txtEnsure that you export OPENAI_API_KEY as part of your environment variables:
export OPENAI_API_KEY=<your key here>Refer to main.ipynb for a detailed tutorial on how to use the framework, covering Evaluation, Optimization and the Actor-Critic Loop
CodeQUEST is licensed under the Apache-2.0 License. See the LICENSE file for details.
Open Source @ JPMorganChase