E2B is a secure, sandboxed environment where you can run arbitrary code.

To build this integration, you just need to replace Open Interpreter’s python (which runs locally) with a python that runs on E2B.

First, get an API key here, and set it:

import os
os.environ["E2B_API_KEY"] = "<your_api_key_here>"

Then, define a custom language for Open Interpreter. The class name doesn’t matter, but we’ll call it PythonE2B:

import e2b

class PythonE2B:
    """
    This class contains all requirements for being a custom language in Open Interpreter:

    - name (an attribute)
    - run (a method)
    - stop (a method)
    - terminate (a method)

    Here, we'll use E2B to power the `run` method.
    """

    # This is the name that will appear to the LLM.
    name = "python"

    # Optionally, you can append some information about this language to the system message:
    system_message = "# Follow this rule: Every Python code block MUST contain at least one print statement."

    # (E2B isn't a Jupyter Notebook, so we added ^ this so it would print things,
    # instead of putting variables at the end of code blocks, which is a Jupyter thing.)

    def run(self, code):
        """Generator that yields a dictionary in LMC Format."""

        # Run the code on E2B
        stdout, stderr = e2b.run_code('Python3', code)

        # Yield the output
        yield {
            "type": "console", "format": "output",
            "content": stdout + stderr # We combined these arbitrarily. Yield anything you'd like!
        }

    def stop(self):
        """Stops the code."""
        # Not needed here, because e2b.run_code isn't stateful.
        pass

    def terminate(self):
        """Terminates the entire process."""
        # Not needed here, because e2b.run_code isn't stateful.
        pass

# (Tip: Do this before adding/removing languages, otherwise OI might retain the state of previous languages:)
interpreter.computer.terminate()

# Give Open Interpreter its languages. This will only let it run PythonE2B:
interpreter.computer.languages = [PythonE2B]

# Try it out!
interpreter.chat("What's 349808*38490739?")
LMC MessagesDocker