Simon Willison’s Weblog

Subscribe

open-interpreter (via) This "natural language interface for computers" open source ChatGPT Code Interpreter alternative has been around for a while, but today I finally got around to trying it out.

Here's how I ran it (without first installing anything) using uv:

uvx --from open-interpreter interpreter

The default mode asks you for an OpenAI API key so it can use gpt-4o - there are a multitude of other options, including the ability to use local models with interpreter --local.

It runs in your terminal and works by generating Python code to help answer your questions, asking your permission to run it and then executing it directly on your computer.

I pasted in an API key and then prompted it with this:

find largest files on my desktop

Would you like to run this code? (y/n)  - shows a chunk of Python code and the output - answers: Here are the largest files on your Desktop:     1 Screen Recording 2024-04-28 at 10.37.20 AM.mov - 4.06 GB     2 Gergely-Orosz-podcast.mp4 - 1.18 GB   3 Descript Recording 2023-12-01 09.58.25.696/screen.mp4 - 1.01 GB    4 Screen Recording 2024-04-28 at 11.03.15 AM.mov - 355.89 MB     5 Screen Recording 2024-06-19 at 8.37.57 PM.mov - 289.76 MB  If you need more information or further assistance, feel free to ask!

Here's the full transcript.

Since code is run directly on your machine there are all sorts of ways things could go wrong if you don't carefully review the generated code before hitting "y". The team have an experimental safe mode in development which works by scanning generated code with semgrep. I'm not convinced by that approach, I think executing code in a sandbox would be a much more robust solution here - but sandboxing Python is still a very difficult problem.