Executing Code in E2B Sandbox
I guess this is not exactly a review because I tested only a few features of the E2B Sandbox. I was just curious how to implement a solution that interprets LLM-generated Python code and returns its result in a custom environment. By now, I have generated chat responses using LLMs in text format, so it would be a new experience if I could generate not just the code but also invoke its return value.
Although my original plan was to implement this inside a web application and write a post about that, I would rather choose PowerShell on desktop for now.
The terminal looks usually boring, but using the Rich Python library can enhance the experience. By the way, I am also using Rich for the first time; previously, I used Colorama to colorize the black and white console window.
What is Sandbox?
Sandbox originally refers to a safe playground for children, yet the technical concept is to provide a safe environment by evaluating code without the risk of prompt injection.
Setting up E2B Sandbox
Prerequisites
- Python >=3.11.5
- Claude API-key
I will use E2B’s sandbox. You can learn from them and their goals here.
After signing up find and copy the API-key at dashboard and paste it to an .env file.
As for the LLM, I will use Claude Anthropic API here. You can get API-key after registartion or use your own from other provider.
# .env
E2B_API_KEY=e2b_***
ANTHROPIC_API_KEY=***
First step is to install the E2B package and other dependencies of our project.
pip install e2b-code-interpreter python-dotenv anthropic rich
For the Sandbox persistency.
pip install e2b-code-interpreter==1.2.0b5
File structure
|- e2b_project\
| |- .venv\
| |- .env
| |- main.py
| |- utils.py