#Sandbox

The sandbox provides an isolated workspace where agents can execute code, manage files, and produce artifacts safely.

#Overview

When an agent needs to run code — data analysis, file processing, generating visualizations — it uses the sandbox. The sandbox provides:

• Isolated execution environment
• File system for reading and writing files
• Support for multiple languages (Python, JavaScript, shell)
• Persistent workspace within a session

#Code Execution

Agents execute code using the sandbox tools:

# The agent can run Python code like this
import pandas as pd
import matplotlib.pyplot as plt

df = pd.read_csv('data.csv')
df.plot(kind='bar', x='category', y='value')
plt.savefig('chart.png')

The execution environment includes common packages. Agents can install additional packages when the selected execution target allows it.

#Supported Languages

• Python — Full CPython environment with scientific computing packages
• JavaScript / Node.js — Server-side JavaScript execution
• Shell — Bash commands for system operations

#File Operations

Agents can read and write files in the sandbox:

Files persist within a sandbox session. Agents can write code to a file, execute it, and read the output — all within the same conversation.

#IDE Preview

The sandbox includes a web-based IDE (code-server) that you can access directly from the SciLaxy interface. This lets you:

• Browse and edit files created by the agent
• Run commands in a terminal
• Preview generated HTML, images, and other outputs

The IDE preview is accessible from the sandbox panel in the chat interface.

#Availability & Lifecycle

Sandbox access follows the execution target assigned to the current Project. It is available only while that target is online and ready. If the target becomes unavailable, SciLaxy reports the interruption and the action required to recover; it does not silently move execution to another device or a hosted fallback.