Running Commands

The Sandbox SDK provides the run_command() method to execute shell commands in your isolated sandbox environment.

​

The run_command() Method

def run_command(
    command: str
) -> Dict[str, Any]

• command (str): The shell command to execute in the sandbox
• Returns a dictionary with:
  • stdout (str): Standard output of the command
  • stderr (str): Standard error output
  • exit_code (int): Exit code of the process

​

Basic Usage

# List files in the sandbox
result = sdk.run_command("ls -la")
print(result["stdout"])

# Install Python packages
sdk.run_command("pip install pandas numpy matplotlib")

# Run a Python script
sdk.run_command("python3 /home/user/analysis.py")

​

Running Multiple Commands

You can chain commands using && or run them in background:

# Chain commands
sdk.run_command("cd /home/user && mkdir data && ls -la")

# Start a background job
sdk.run_command("python3 -m http.server 8080 &")

​

Handling Results and Errors

result = sdk.run_command("grep TODO *.py")
if result["exit_code"] == 0:
    print("Matches found:")
    print(result["stdout"])
else:
    print("No matches or error:")
    print(result["stderr"])

​

Advanced Examples

​

Running Interactive or Long-Running Processes

For interactive sessions or long-running jobs, you can run in background:

# Launch a live server in the sandbox
sdk.run_command(
    "cd /home/user/project && uvicorn app:app --reload &"
)

​

Installing System Packages

# Update and install system packages
sdk.run_command("apt-get update && apt-get install -y curl vim")

​

Managing Working Directory

All commands are executed in the sandbox root by default. Use cd to change directories:

sdk.run_command("cd /home/user/project && pytest --maxfail=1 --disable-warnings")

​

Best Practices

• Use absolute paths (/home/user/...) to avoid ambiguity
• Chain commands when appropriate to reduce round-trips
• Run long-running processes in background using &
• Always check exit_code to handle errors programmatically