Sandbox Fallback Execution 5eda7b

execute_code_sandbox(code="your_python_code_here")

Step 2: Detect Failure and Switch to Fallback

When sandbox execution fails with initialization errors, switch to the fallback pattern:

1. Write the Python script to disk using write_file:

   • Choose a descriptive filename (e.g., generate_pdf.py, process_data.py)
   • Include the complete Python code with all necessary imports

2. Execute via shell using run_shell:

   • Run the script with python or python3
   • Capture stdout/stderr for verification

Step 3: Example Implementation

# Write the script to disk
write_file(
    path="generate_pdf.py",
    content="""
from reportlab.lib.pagesizes import letter
from reportlab.pdfgen import canvas

def create_pdf(filename, content):
    c = canvas.Canvas(filename, pagesize=letter)
    c.drawString(100, 750, content)
    c.save()

create_pdf('output.pdf', 'Hello World')
"""
)

# Execute the script via shell
run_shell(command="python generate_pdf.py")

Step 4: Handle Dependencies

If the script requires external packages:

# Install dependencies first
run_shell(command="pip install reportlab pillow")

# Then execute the script
run_shell(command="python generate_pdf.py")

Step 5: Verify Output

After execution, verify the output was created:

# Check if file was created
run_shell(command="ls -la output.pdf")

# Optionally read the file to confirm
read_file(file_path="output.pdf", filetype="pdf")

Best Practices

1. Keep scripts self-contained: Include all imports and logic in the written file
2. Use descriptive filenames: Make it clear what each script does
3. Clean up when done: Remove temporary scripts if not needed for debugging
4. Capture errors: Always check stdout/stderr from run_shell for debugging
5. Handle paths carefully: Use relative paths or absolute paths consistently

Complete Example Pattern

# Primary: Try sandbox execution