k2-sandbox 0.1.1

Python SDK for running LLM-generated code in isolated Docker container environments

K2 Sandbox

K2 Sandbox is a Python SDK for running LLM-generated code in isolated Docker container environments. It provides a secure way to execute untrusted code with filesystem access, network capabilities, and stateful execution.

Features

• Secure code execution in isolated Docker containers
• Support for Python, JavaScript, and TypeScript execution
• Stateful execution (variable persistence between runs)
• File system operations
• Process management
• Terminal interaction

Installation

pip install k2-sandbox

Quick Start

from k2_sandbox import Sandbox

# Optionally, set the API base URL globally if needed
# This is useful if your API server is not at the default http://localhost:3000
# and you don't want to set it via environment variable or for every sandbox instance.
Sandbox.set_global_default_api_base_url("https://your-k2-api-server.com")

# Create and use a code interpreter sandbox
with Sandbox.create_code_interpreter() as sandbox:
    execution = sandbox.run_code("x = 41; x + 1")
    print(f"Result: {execution.text}")  # Output: 42

Documentation

For detailed documentation, see the API Reference.

License

MIT