[REQUEST] Add Local Code Sandbox support

[phanisarman]

Feature Request

Is your feature request related to a problem? Please describe.

Python and JavaScript with npm packages currently require E2B (external cloud service), making them unavailable in air-gapped/restricted environments. This blocks enterprise deployments with compliance requirements that prohibit outbound traffic.

Describe the solution you'd like

Implement a local sandbox environment similar to DifySandbox with these capabilities:

Core Features:

1. Multi-language support: Python and JavaScript/Node.js
2. Dependency management:
   • Python: requirements.txt or inline declarations
   • JavaScript: package.json or inline imports
3. Security isolation:
   • Seccomp-based system call filtering (whitelist policy)
   • File system isolation
   • Network isolation via Docker network/proxy
4. Pre-configured packages: Common libraries (numpy, pandas, requests)
5. Custom dependencies: Users can add packages via config

Configuration:

LOCAL_SANDBOX_ENABLED=true
LOCAL_SANDBOX_PYTHON_PACKAGES=numpy,pandas,requests,beautifulsoup4

Implementation approach (following Dify's architecture):

• Isolated Docker containers with Seccomp security profiles
• Dependency caching to avoid repeated installations
• Network proxy for controlled external access (optional)
• Resource limits (CPU, memory, timeout)

Describe alternatives you've considered

1. E2B only - Requires internet access, not viable for air-gapped
2. Basic execution without packages - Too limited for real-world use
3. Dify solves this with DifySandbox - a local, secure code execution environment that runs Python and Node.js with custom dependencies in self-hosted deployments.

Additional context

Reference Implementation:

• Dify's DifySandbox: https://github.com/langgenius/dify-sandbox
  • Supports Python 3.10+ and Node.js locally
  • Pre-installed dependencies with custom additions
  • Battle-tested in production self-hosted deployments
  • Built on Seccomp for system-level security

Use cases this enables:

• Data processing with pandas/numpy
• API integrations with custom libraries
• ML inference with scikit-learn/transformers
• Web scraping with beautifulsoup4
• All in fully air-gapped environments

Impact:

• Enables government, healthcare, financial deployments
• ~30-40% of enterprise self-hosted market has network restrictions
• Feature parity with Dify, n8n etc.

[Xlaudius]

I'm locally hosting though Docker in AWS, and this feature would be awesome to have, since I'm trying to move from n8n.

I cannot seem to even get crypto working to generate UUIDs

[waleedlatif1]

@phanisarman Great suggestion! We will get to work on this. By default, it executes the code locally in a VM so it doesn't have round-trip latency or need to leave your environment, but unfortunately only fetch is baked into this VM and it only supports JS/TS. We can add an environment variable that allows you to specify what packages you want to include, and work on python code execution without ever leaving your machine.