How To Install GPT-Engineer👨💻- Build ENTIRE Apps With One Prompt! (Quick Setup Guide)
Introduction to GPT Engineer
In this section, the speaker introduces a new AI coding partner called GPT Engineer and demonstrates its capabilities. The speaker also explains how to install and use it.
Installing GPT Engineer
- Clone the repo from GPT Engineers GitHub page.
- Install conda to avoid any Python and module versioning issues.
- Create a new conda environment using a recent version of python.
- Activate the environment and install all necessary modules using
python -m pip install -r requirements.txt.
Using GPT Engineer
- Open VS Code and open the specific directory for GPT Engineer.
- Define the project by describing it in the main prompt file located in the example directory.
- Get an OpenAI API key and enter it into terminal using
export open AI API key=<API KEY>.
- Run the main file using
python main.py examplewhich references the folder that contains that main prompt file.
Clarifying Questions
- When running the program, clarifying questions will be asked to ensure accuracy.
Defining Project Prompt
In this section, we learn how to define a project prompt for GPT Engineer.
Defining Project Prompt
- Describe your desired project in as much detail as possible in the main prompt file located in the example directory.
- Be prepared to answer clarifying questions when running the program.
Running Main File
In this section, we learn how to run the main file for our defined project prompt.
Running Main File
- Use
python main.py examplecommand to run your defined project prompt.
- Be prepared to answer clarifying questions when running the program.
Building the Project
In this section, the speaker shows how to build a project using GPT-3.
Creating Files
- The speaker speeds up the video while showing that files are being written out.
- After the files are created, they can be found in the example folder under workspace.
Main File
- The main file is called "main.pi" and contains everything needed to start the project, including classes for snake, game, and game view.
Conclusion
- Although GPT-3 cannot currently read an existing code base and iterate on it, it can write entire projects with a single prompt. The speaker encourages viewers to get detailed in their prompt descriptions.