envforge-agent

Standalone CLI diagnostic agent for the EnvForge platform.

Inspects your local ML environment and reports what's installed, what's compatible, and what's broken — without requiring a network connection.

Install

pip install envforge-agent

Commands

# Inspect your environment (output to terminal)
envforge diagnose

# Save report to JSON file
envforge diagnose --output report.json

# Send report to EnvForge API for compatibility analysis
envforge diagnose --send --api-url https://api.envforge.dev

# Check if a specific profile is compatible with your system
envforge verify --profile pytorch-cuda

# Generate a repair script from a saved diagnostic report
envforge fix --report report.json

What it detects

Output format

All commands output DiagnosticReport JSON compatible with POST /api/v1/diagnose.

{
  "agent_version": "1.0.0",
  "os": { "name": "Ubuntu 22.04", "version": "22.04", "architecture": "x86_64" },
  "cpu": { "brand": "Intel Core i9-13900K", "cores": 24, "threads": 32 },
  "ram": { "total_gb": 64.0, "available_gb": 48.2 },
  "gpus": [{ "name": "NVIDIA RTX 4090", "vram_gb": 24.0, "driver_version": "535.54", "index": 0 }],
  "cuda": { "version": "12.1", "toolkit_path": "/usr/local/cuda-12.1", "cudnn_version": "8.9.0" },
  "python_installations": [{ "version": "3.11.9", "path": "/usr/bin/python3.11", "is_venv": false }],
  "active_python": { "version": "3.11.9", "path": "/usr/bin/python3.11", "is_venv": false }
}

Platform support

Troubleshooting

For common CLI, API, and environment issues, see:

../TROUBLESHOOTING.md