just-agents-coding 0.5.5

Just Agents - Coding Components

just-agents-coding

A submodule of just-agents focused on code generation and execution capabilities.

Overview

just-agents-coding provides secure code execution environments and tools for LLM-powered coding agents. It enables safe code generation and execution through containerized environments, protecting your system while allowing AI agents to write and test code.

Key Features

• 🔒 Secure code execution through isolated containers
• 🐳 Multiple specialized containers:
  • Standard sandbox for general Python code
  • Biosandbox for bioinformatics tasks
  • Websandbox for web-related code
• 📁 Simple I/O management with mounted /input and /output directories
• 🔍 Code execution monitoring and logging

Quick Start

pip install just-agents-coding

Usage Example

from just_agents.base_agent import BaseAgent
from dotenv import load_dotenv

# Load environment variables
load_dotenv()

# Initialize agent from configuration
agent = BaseAgent.from_yaml("SimpleCodeAgent", file_path="path/to/coding_agents.yaml")

# Execute code through the agent
result = agent.query("""
Get FGF2 human protein sequence from uniprot using biopython.
As a result, return only the sequence
""")

print(result)

Container Types

Standard Sandbox

• General Python code execution
• Basic Python packages pre-installed
• Isolated from host system

Biosandbox

• Specialized for bioinformatics tasks
• Includes common bio packages (Biopython, etc.)
• Safe handling of biological data

Websandbox

• Web development and testing
• Network access controls
• Common web frameworks available

Security Features

• Root access disabled in containers
• Resource usage limits
• Network isolation
• Temporary file system
• Controlled package installation

Documentation

For more detailed documentation and examples, visit:

• Basic Usage Tutorial