impllmentors is a POC project demonstrating the emerging approach of structured and flow-controlled LLM agents (as opposed to freely roaming agents as in AutoGPT and BabyAGI).
It is humble, and can only implement simple python modules. However simple, it also:
- Has well-defined structure and steps, avoids running in circles.
- Asks for human feedback between steps.
- Writes unit tests.
- Runs the unit tests, and if they fail, tries to fix the code/
- Clone the repository:
git clone https://github.com/stavshamir/impllmentors.git - Run
pip install -f requirements.txt - Set an
OPENAI_API_KEYenvironment variable - Run
python runner.py
- The generated code and tests are written to the directory from which you run the script.
- You can view the written files and edit them between steps to help the AI.
- After the unit tests are written, the AI will not attempt to fix them in case they fail.
- This uses gpt-3.5-turbo, as I (and many others) still don't have access to the gpt-4 api. This means that hallucinations still happen quite often.
(Sorry for the low quality)
Feel free to suggest improvements and provide PRs!