Your own ChatGPT Code Interpreter, no Plus required

What it does

gpt-code-ui is a web interface that wires an OpenAI model to a local Python kernel. You type a request in natural language, the model writes Python code, and the UI immediately executes it on your machine. It supports file uploads and downloads, so you can feed it files and ask for plots, cleaning, or analysis without leaving the browser.

The interesting bit

The project recreates the full feedback loop of OpenAI’s Code Interpreter—generate, execute, display—using a local Python kernel instead of a hosted sandbox. What sets it apart from a simple chat wrapper is the tight coupling between the LLM and live code execution, though it still relies entirely on OpenAI’s API for the reasoning layer.

Key highlights

• Runs generated Python locally in a live kernel, not just as static text output
• Supports GPT-3.5 and GPT-4 switching from the UI
• File upload and download for working with real datasets
• Context awareness across conversation turns
• Configurable OpenAI base URL, including Azure OpenAI Services support

Caveats

• Model selection in the UI is currently unsupported when using Azure OpenAI Services
• The README recommends manually installing pinned data-science dependencies (numpy, pandas, matplotlib, and others) in the execution environment; they are not bundled with the package
• Requires an OpenAI API key or Azure equivalent; it is not a local LLM runner

Verdict

Worth a spin if you want the Code Interpreter experience on your own hardware or need to process sensitive files without shipping them to ChatGPT’s hosted environment. Skip it if you are looking for a fully offline, open-weights alternative that never touches OpenAI’s API.