Skip to content

Latest commit

 

History

History
28 lines (17 loc) · 2.15 KB

CODE-INTERPRETER.md

File metadata and controls

28 lines (17 loc) · 2.15 KB
Raw

Code Interpreter

This bot supports a full fledged version of Code Interpreter, where code in various languages can be executed directly in Discord. You can even install Python and system packages. Python is the preferred language for code interpreter, although it will still work relatively nicely with other popular languages.

To get started with code interpreter, you need an E2B API key. You can find an E2B API key & more info on FoundryLabs here.

E2B_API_KEY="...."

Like above, add the E2B API key to your .env file. E2B is a cloud-based isolated execution environment, so that you can run code safely and have a containerized environment to install packages and execute code, etc in.

Afterwards, to use code interpreter, simply use /code chat.

When you begin a code interpreter instance, a new isolated environment to run your code is automatically created. Inside your chat, you can ask GPT to install Python or system packages into this environment, and ask GPT to run any sort of python (and other language) code within it as well. Unlike ChatGPT's code interpreter / advanced data analysis, this also has access to the internet so you can work with code that uses the network as well.

As per the E2B documentation, the free tier provides the following resources to a sandboxed code instance:

  • 2 vCPU
  • 512 MB RAM
  • 1 GB free disk storage
  • 24 hours max sandbox session length

Given that it is possible to maintain and use a thread over multiple days with GPTDiscord, exceeding the instance limits of your E2B API key will cause the thread to break and not reply. In this case, you will need to end & restart the thread to continue conversing.

When ChatGPT executes code, sometimes it will create files (especially if you ask it to within your instructions) in a folder called artifacts. If files are created, they can be downloaded from the "Download Artifacts" button that will pop up after the code is executed.

Sometimes, ChatGPT will create files that are not placed in the upstream artifacts folder. To remedy this, simply ask it to ensure all files it makes are within artifacts.