Sandbox MCP

# Python Sandbox: FAQ and Limitations ## Overview The Python Sandbox provides a secure, isolated environment for executing Python code with advanced features including artifact management, web app hosting, and Manim animation support. ## Frequently Asked Questions ### 1. Environment and Setup **Q: What Python version does the sandbox use?** A: The sandbox runs Python 3.11+ in a dedicated virtual environment located at `/home/usr/sandbox/.venv`. **Q: What libraries are pre-installed?** A: Core libraries include: - `numpy` - Numerical computing - `matplotlib` - Plotting and visualization - `pandas` - Data manipulation - `PIL/Pillow` - Image processing - `manim` - Mathematical animations - `flask` - Web application framework - `streamlit` - Data app framework - `jupyter` - Interactive computing **Q: Can I install additional packages?** A: Yes, while all commonly needed packages are pre-installed, the install commands can be used to safely install packaged that do not require sudo permissions. ### 2. Code Execution **Q: How do I properly format code for execution?** A: Follow these guidelines: ```python # ✅ Correct - Use forward slashes for paths plt.savefig('/artifacts/plot.png') # ❌ Incorrect - Avoid backslashes plt.savefig('\\artifacts\\plot.png') # ✅ Correct - Use raw strings for Windows-style paths if needed path = r'C:\Users\data\file.txt' ``` **Q: Why do I get syntax errors with my code?** A: Common issues: - Trailing backslashes in strings - Incorrect path separators - Missing imports - Indentation errors **Q: How long can code execute?** A: Default timeout is 30 seconds for Python code and 10 seconds for shell commands. ### 3. Artifact Management **Q: Where are artifacts stored?** A: Artifacts are stored in session-specific directories: - Base path: `/home/usr/sandbox/sessions/{session_id}/artifacts/` - Subdirectories: `plots/`, `images/`, `videos/`, `data/`, `manim/` **Q: What file types are supported as artifacts?** A: Supported formats: - **Images**: PNG, JPG, JPEG, GIF, BMP, SVG - **Videos**: MP4, AVI, MOV, WEBM - **Data**: CSV, JSON, XML, YAML, HDF5 - **Documents**: PDF, TXT, MD - **Code**: PY, HTML, CSS, JS **Q: How do I ensure my artifacts are saved?** A: Use absolute paths to the artifacts directory: ```python import matplotlib.pyplot as plt import os # Create the artifacts directory if it doesn't exist os.makedirs('/artifacts/plots', exist_ok=True) # Save your plot plt.savefig('/artifacts/plots/my_plot.png') ``` ### 4. Web Applications **Q: How do I create a web app?** A: Use Flask or Streamlit: ```python # Flask example from flask import Flask app = Flask(__name__) @app.route('/') def home(): return "Hello from Sandbox!" if __name__ == '__main__': app.run(host='0.0.0.0', port=8000) ``` **Q: How do I access my web app?** A: Web apps are accessible at `http://127.0.0.1:8000` (or assigned port) from your local machine. ### 5. Manim Animations **Q: How do I create Manim animations?** A: Use the pre-compiled examples or create custom scenes: ```python from manim import * class MyScene(Scene): def construct(self): circle = Circle() self.play(Create(circle)) self.wait(1) ``` **Q: Why don't my Manim animations render?** A: Common issues: - Incorrect scene class structure - Missing `construct` method - Syntax errors in animation code - Memory limitations for complex scenes ### 6. Interactive Features **Q: What REPL features are available?** A: Enhanced REPL includes: - Command history - Tab completion (basic) - Custom commands: `artifacts`, `stats`, `help` - Session persistence - Colored output **Q: How do I access previous execution results?** A: Use the history command in REPL: ```python history() # Show recent executions stats() # Show performance statistics ``` ## Limitations and Restrictions ### 1. Filesystem Access **Restrictions:** - ❌ No access to system directories outside sandbox - ❌ Cannot modify system files or configurations - ❌ Limited to session-specific artifact directories - ❌ No network access to external APIs (security) **Allowed:** - ✅ Read/write to artifact directories - ✅ Create temporary files in session space - ✅ Access pre-installed library data ### 2. Resource Limitations **Memory:** - Maximum 1GB RAM per session - Automatic cleanup of large objects - Session isolation prevents memory leaks **CPU:** - Single-threaded execution - 30-second timeout for Python code - 10-second timeout for shell commands **Storage:** - 100MB maximum per session - Automatic cleanup of old sessions - No persistent storage between sessions ### 3. Network and External Access **Restrictions:** - ❌ No internet access for security - ❌ Cannot access external databases - ❌ No external API calls - ❌ Cannot send emails or notifications **Allowed:** - ✅ Local web server hosting - ✅ Inter-process communication within sandbox - ✅ File-based data exchange ### 4. Package Management **Restrictions:** - ❌ Cannot install new packages (`pip install`) - ❌ Cannot modify package configurations - ❌ No access to package source code modification **Alternatives:** - ✅ All common packages pre-installed - ✅ Multiple versions of key libraries available - ✅ Request additional packages for future updates ### 5. Shell Access **Restrictions:** - ❌ No system administration commands - ❌ Cannot modify system services - ❌ Limited to sandbox directory operations - ❌ No access to sensitive system information **Allowed:** - ✅ File operations within sandbox - ✅ Basic utilities (ls, cat, grep) - ✅ Python script execution - ✅ Text processing commands ## Best Practices ### 1. Code Organization ```python # Structure your code clearly import numpy as np import matplotlib.pyplot as plt import os # Set up artifact directory artifact_dir = '/artifacts/plots' os.makedirs(artifact_dir, exist_ok=True) # Your main code data = np.random.randn(1000) plt.hist(data, bins=50) plt.title('Random Data Distribution') # Save with absolute path plt.savefig(f'{artifact_dir}/histogram.png') plt.show() ``` ### 2. Error Handling ```python try: # Your code here result = some_operation() except Exception as e: print(f"Error: {e}") # Log error for debugging with open('/artifacts/error_log.txt', 'a') as f: f.write(f"Error at {datetime.now()}: {e}\n") ``` ### 3. Resource Management ```python # Clean up large objects import gc # After processing large data del large_data_structure gc.collect() # Use context managers for files with open('/artifacts/data.csv', 'w') as f: f.write(data_string) ``` ### 4. Session Management ```python # Save important session data session_data = { 'variables': list(globals().keys()), 'artifacts': os.listdir('/artifacts'), 'timestamp': datetime.now().isoformat() } with open('/artifacts/session_info.json', 'w') as f: json.dump(session_data, f, indent=2) ``` ## Troubleshooting Guide ### Common Issues and Solutions **1. "No artifacts found" error** - Check artifact directory exists: `os.makedirs('/artifacts', exist_ok=True)` - Use absolute paths: `/artifacts/filename.ext` - Verify file was actually created: `os.path.exists('/artifacts/filename.ext')` **2. "Syntax error" in code execution** - Remove trailing backslashes from strings - Check indentation consistency - Verify all imports are included **3. "Web app not accessible"** - Ensure host is set to '0.0.0.0' - Check port availability (8000-8010 range) - Verify Flask/Streamlit is running **4. "Manim animation failed"** - Check scene class structure - Verify `construct` method exists - Ensure proper Manim imports **5. "Memory limit exceeded"** - Reduce data size or complexity - Use generators instead of lists for large datasets - Clean up unused variables with `del` ## Getting Help For additional support: 1. Use the `help()` command in REPL 2. Check example code in `/examples/` directory 3. Review error logs in `/artifacts/logs/` 4. Test with minimal code examples first ## Security Notes The sandbox environment is designed for educational and development purposes with the following security measures: - Process isolation - Filesystem restrictions - Network isolation - Resource limitations - Automatic session cleanup These limitations ensure safe execution while providing powerful development capabilities within the restricted environment.