Code Interpreter

Safe and sandboxed Python runtime available to the LLM

Overview

Code Interpreter example

Example run where the LLM generates code to run data analysis on a business report.

The ability to execute code unlocks the ability for the LLMs to do things such as:

• Performing larger calculations accurately
• Running data analysis on provided files
• Generating or modifying data/files

Note: The code interpreter is built-in functionality available to all deployments without any configuration needed.

Features

• Python Runtime: A secure Python runtime with all dangerous functionality removed (such as accessing the network, directories outside of the specified, etc.)
• Libraries: Comes with a set of libraries such as numpy, pandas, scipy, matplotlib, and more.
• File Input: Ability to pass arbitrary file types along with the code to run against it.
• File Output: Files can be created and provided back to the user.
• STDIN/STDOUT Capture: Ability to show the user any outputs of the programs executed.
• Graph Rendering: The Onyx Chat UI is able to render returned graphs and other visualization for the user.

Usage

The Code Interpreter does not need to be invoked explicitly, the LLM will determine when to use it depending on the user query.

It can also be attached to custom Agents, giving the LLM the option to use it as needed.