This project provides secure code execution capabilities for Open WebUI, enabling users to run code directly within the interface or grant LLMs the ability to execute code autonomously. It targets Open WebUI users and developers seeking to integrate reliable code execution into their AI workflows, offering enhanced interactivity and LLM capabilities.

How It Works

The project leverages gVisor for secure sandboxing, mirroring the approach used by ChatGPT. It offers two primary components: a "Code execution function" that allows users to manually run code blocks via a button, displaying output in the UI and feeding it back to the LLM; and a "Code execution tool" that empowers LLMs to autonomously execute code, with outputs visible only to the LLM for its decision-making process.

Quick Start & Requirements

• Installation: Requires prior setup of Open WebUI for sandboxing. Code is then added as a "Function" or "Tool" within Open WebUI's Workspace settings.
• Prerequisites: Open WebUI with sandboxing enabled.
• Resources: No specific hardware requirements mentioned beyond running Open WebUI.
• Links: Open WebUI Sandboxing Setup (Note: The provided README links to a general Open WebUI setup, not a specific sandboxing guide).

Highlighted Details

• Integrates directly into Open WebUI as either a user-triggered function or an LLM-accessible tool.
• Utilizes gVisor for robust, secure sandboxing of code execution.
• Enables LLMs to perform dynamic tasks like retrieving current information or executing complex computations.

Maintenance & Community

• Developed by EtiennePerot.
• No specific community links (Discord, Slack) or roadmap details are provided in the README.

Licensing & Compatibility

• The README does not explicitly state a license.

Limitations & Caveats

• The project's functionality is entirely dependent on a correctly configured Open WebUI environment with sandboxing enabled.
• The README does not specify supported programming languages for code execution.