A quick reference guide for using the sjsujetsontool utility on NVIDIA Jetson devices. Full tutorial: 00_sjsujetsontool_guide.md
📋 Commands by Category
📥 Installation
| Command |
Description |
curl -fsSL https://raw.githubusercontent.com/lkk688/edgeAI/main/jetson/install_sjsujetsontool.sh \| bash |
Install (no sudo) |
sjsujetsontool update |
Update container & script |
sjsujetsontool update-container |
Update container only |
sjsujetsontool update-script |
Update script only |
🔍 Basic Commands
| Command |
Description |
sjsujetsontool list |
Show all commands |
sjsujetsontool version |
Show versions |
sjsujetsontool status |
Show container & GPU stats |
sjsujetsontool debug |
Run diagnostics |
sjsujetsontool setup-check |
Check /Developer folder and clone/pull edgeAI git repo |
🐳 Container Management
| Command |
Description |
sjsujetsontool shell |
Enter container shell |
exit |
Exit shell (container keeps running) |
sjsujetsontool stop |
Stop container completely |
🐍 Python & Jupyter
| Command |
Description |
sjsujetsontool run /path/to/script.py |
Run Python script in container |
sjsujetsontool jupyter |
Launch JupyterLab (port 8888) |
🧠 Ollama LLM Commands
| Command |
Description |
sjsujetsontool ollama-serve |
Start server (shortcut for ollama serve, port 11434) |
sjsujetsontool ollama-run gemma4 |
Run model interactively (shortcut for ollama run) |
sjsujetsontool ollama list |
List installed models |
sjsujetsontool ollama pull gemma4 |
Download model |
sjsujetsontool ollama delete gemma4 |
Remove model |
sjsujetsontool ollama status |
Check server status |
sjsujetsontool ollama ask "What is NVIDIA Jetson?" |
Ask question (auto-pulls) |
🔬 Llama.cpp & Gemma 4 E2B (VLM) Commands
| Command |
Description |
sjsujetsontool llama |
Start Gemma 4 E2B llama-server (port 8080) |
sjsujetsontool llama-cli -p "prompt" |
Run Gemma 4 E2B llama-cli text query |
sjsujetsontool llama-cli --image /Developer/LoveSJ-hero-4.png -p "Describe" |
Run Gemma 4 E2B llama-cli image query |
curl http://localhost:8080/v1/chat/completions -d '...' |
Query OpenAI-compatible Chat completions API |
http://localhost:8080 |
Access web UI dashboard |
🚀 vLLM Speculative Decoding Commands
| Command |
Description |
sjsujetsontool vllm [model] |
Start vLLM serve engine (defaults to Qwen3-8B-speculator on port 8000) |
curl http://localhost:8000/v1/chat/completions -d '...' |
Query vLLM OpenAI-compatible REST API |
🔑 NVIDIA Build API / NGC Cloud Inference
| Command |
Description |
sjsujetsontool setup-nvapi |
Setup NVIDIA NGC API Key in .env.local & perform validation test |
sjsujetsontool nv-chat |
Start interactive chat with curated model selection (Llama 3.x, Nemotron) |
sjsujetsontool nv-chat "prompt" |
Run single prompt query with default model and performance metrics |
sjsujetsontool nv-chat -p "prompt" -m [model] |
Run single query with a customized model (e.g. Nemotron Omni) |
Mounted Paths
- Host directories mounted in container:
/Developer → /Developer in container/Developer/models → /models in container
SSH Connectivity
- Connect via mDNS hostname:
ssh username@jetson-hostname.local
- Example:
ssh sjsujetson@sjsujetson-01.local
- For X11 forwarding:
ssh -X sjsujetson@sjsujetson-01.local
⚠️ Safety Tips
- Power Supply: Use a 5A USB-C adapter or official barrel jack for power stability
- Containers: Always stop containers with
sjsujetsontool stop before unplugging
- SSD Cloning: Change hostname and machine-id after cloning to prevent network conflicts
- SSH Security: Only install SSH keys from trusted GitHub accounts
- Disk Cleanup: Remove cache and large datasets before creating system images