A Jupyter notebook extension that provides AI-powered feedback on your code execution and outputs. Per defaut, it uses the gpt-4.1-nano to analyze your code and provide suggestions for improvements.
Caution
unprompted is a research tool intended to streamline data analysis experience by identifying issues in code early. It is under development and certainly not perfect. Under the hood it uses large language models, which may do mistakes. Read feedback carefully and critically before following it.
When using the OpenAI, Github Models or any other LLM service provider with unprompted, you are bound to the terms of service of the respective companies or organizations. The code you enter and its output are transferred to their servers and may be processed and stored there. Make sure to not submit any sensitive, confidential or personal data. Also using these services may cost money.
- Python 3.9+
- Access to an OpenAI-API compatible LLM-server. Per default, unprompted uses OpenAI's commercial services.
- Alternatively, you can use Ollama. In this case, an NVidia GPU with > 4 GB memory is recommended.
Install the required Python packages:
pip install unpromptedIn case of local ollama:
- Follow the installation instructions at ollama.com
- Start the Ollama server (this happens automatically on Windows)
- Download the a compatible model:
ollama pull gemma3:4bModels that also work technically. Quality wasn't measured yet.
In your Jupyter notebook, import the extension full example:
import unpromptedIf you want to use ollama, configure the environment variables UNPROMPTED_MODEL, UNPROMPTED_LLM_URL and `UNPROMPTED_API_KEY in case of a remote server. Check out the corresponding example notebook for details.