deploy-live-vlm-webui

name	deploy-live-vlm-webui
description	Deploy Live VLM WebUI on reComputer Jetson for real-time Vision Language Model interaction via webcam. Installs Ollama with llama3.2-vision model and the live-vlm-webui Python package for browser-based VLM streaming and benchmarking. Requires JetPack 6.2 and USB camera.

Deploy Live VLM WebUI on reComputer Jetson

Live VLM WebUI streams your webcam to any VLM for live AI-powered visual analysis. This skill installs Ollama + llama3.2-vision and the WebUI on Jetson.

---

Execution model

Run one phase at a time. After each phase:

• Relay all output to the user.
• If output contains [STOP] → stop, consult the failure decision tree.
• If output ends with [OK] → tell the user "Phase N complete" and proceed.

---

Prerequisites

Requirement	Detail
Hardware	reComputer Super J4012 (Orin NX 16GB) or similar
JetPack	6.2
Camera	USB camera connected to Type-A port
Network	Internet for model download (~7GB)

---

Phase 1 — Install and configure Ollama (~10–15 min)

curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3.2-vision:11b

Verify:

ollama list
# Expected: llama3.2-vision:11b listed

[OK] when model appears in ollama list. [STOP] if download fails or OOM.

---

Phase 2 — Install Live VLM WebUI (~3 min)

sudo apt install -y openssl python3-pip
python3 -m pip install --user live-vlm-webui
echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc
source ~/.bashrc

Verify:

which live-vlm-webui

[OK] when the binary path is printed. [STOP] if pip install fails.

---

Phase 3 — Launch and configure WebUI (~2 min)

live-vlm-webui

Open browser at https://localhost:8090, then configure:

1. VLM API Configuration → select ollama engine → select llama3.2-vision model
2. Camera and App Control → select USB Camera
3. Click Run to start inference

[OK] when inference results appear in the browser.

---

Failure decision tree

Symptom	Action
Ollama install fails	Check internet. Retry: curl -fsSL https://ollama.com/install.sh | sh.
Model pull OOM or killed	16GB RAM minimum for 11b model. Free memory by stopping other processes.
live-vlm-webui: command not found	Ensure ~/.local/bin is in PATH: source ~/.bashrc.
WebUI not accessible at port 8090	Check firewall: sudo ufw allow 8090. Verify process is running.
USB camera not detected in WebUI	Check camera: ls /dev/video*. Replug camera.
Inference very slow	Expected on 16GB devices. Consider using a smaller vision model.

---

Reference files

• references/source.body.md — Full Seeed Wiki tutorial with screenshots, demo video, and configuration details (reference only)