This project is a Python program that allows you to interactively explore your codebase. It's designed to find all programming files in a specified directory and its subdirectories, and then uses OpenAI's GPT-4 model to answer any questions you have about a selected file.
Before you can run this program, you'll need to have the following installed on your machine:
- Python 3.8 or higher
- pip (Python's package installer)
To get started with this project, follow these steps:
- Clone the repository to your local machine:
git clone https://github.com/colesmcintosh/script-sleuth- Navigate to the project directory:
cd code-qa- Install the necessary Python packages:
pip install -r requirements.txt- Create a file in the root project directory named
.envand add your OpenAI API key as an environment variable:
OPENAI_API_KEY=your_openai_api_keyReplace your_openai_api_key with your actual OpenAI API key.
Once you've installed all prerequisites and set up your environment variables, you can run the program with the following command:
python main.pyWhen you run the program, it will first ask you to enter a directory. This should be the root directory of the codebase you want to explore. The program will then find all programming files in that directory and its subdirectories, presenting you with a list of these files.
Next, the program will ask you to enter the number of the file you're interested in. Once you've selected a file, you will be prompted to input a question about the selected file. The program will then use the GPT-4 model to generate an answer to your question and print it out.
The program uses Python's glob and os modules to find all programming files in the specified directory and its subdirectories. The dotenv module is used to load environment variables, including your OpenAI API key, which is used to authenticate with the OpenAI API.
The langchain library is used for language model generation, specifically the ChatOpenAI class for interaction with the OpenAI API, and the ChatPromptTemplate, SystemMessagePromptTemplate, and HumanMessagePromptTemplate classes for prompt creation and formatting.
Script Sleuth comes packed with features that aid in interactive code exploration. Here's a rundown of what's currently available and what could potentially be added in the future.
- Directory-based code navigation: Easily traverse your codebase by inputting a directory path.
- Support for multiple programming languages: The program can process Python, HTML, CSS, Ruby, C, JavaScript, Markdown, CPP, Go, Java, Proto, RST, Rust, Scala, Swift, LaTeX, and Solidity.
- Interactive questioning: You can ask as many questions as you want about a selected file. The program continues to answer until you explicitly decide to exit or switch files.
- Support for backtracking: You can choose a new file from the current directory or go back to choose a new directory without exiting the program.
- AI-powered code analysis: Utilizes the power of OpenAI's GPT-4 model to answer questions about your code.
- HuggingFace model support 🤗: You can use any model from the HuggingFace model hub to answer questions about your code.
Future Considerations:
- File upload support: Instead of inputting a directory, users could directly upload the file they want to analyze.
- User Interface: A simple GUI could improve the user experience, making navigation and interaction even more seamless.
- Integration with code editors: Imagine if this tool could be used directly in your favorite code editor! An extension or plugin could be a future development direction.
- Vector store retrieval to handle larger code bases
Did some testing implementing a vector store but don't really see the necessity for it. If you think it's necessary, feel free to implement it and make a pull request!
As always, Script Sleuth remains a tool for developers, by developers. If you have any suggestions or would like to contribute, don't hesitate to reach out or make a pull request!
Contributions are what make the open source community such an amazing place to learn, inspire, and create. Any contributions you make are greatly appreciated.
Distributed under the MIT License. See LICENSE for more information.
This README was generated using the very program it describes - metaaaaa! 🤩