NVIDIA edge AI system running Ollama + Open WebUI + the PazLabs edge-router — routing queries locally first, escalating to the cloud when confidence drops.
Hardware
Specs
🖥️
Device
NVIDIA Jetson Orin Nano
🧠
RAM
8 GB unified memory CPU + GPU shared
⚡
GPU
Ampere architecture 1024 CUDA cores
🌐
Local IP
192.168.11.50
🐧
OS
Ubuntu 22.04
🐳
Docker
Loading…
🛡️
Argus
Network Monitor systemd service · :8400
💾
microSD
SanDisk Extreme 512GB Up to 245 MB/s read
🚀
NVMe SSD
Samsung 990 Pro 2TB Up to 7,450 MB/s read
Live
System Status
Checking…
Ollama
—
Model Loaded
—
Confidence Threshold
—
Active Cloud APIs
Loading…
Live
Container Status
Checking…
Loading…
Live
Cloud API Status
Checking…
Loading…
Live
Storage
💾
SanDisk Extreme 512GB microSD
Up to 245 MB/s read
25%
127 GB used385 GB free · 512 GB total
⚡
Samsung 990 Pro 2TB NVMe
Up to 7,450 MB/s read
2%
45 GB used1955 GB free · 2000 GB total
Live
Memory
Used1.7 GB (23%)
Cached2 GB (27%)
Free3.7 GB (50%)
How It Works
Architecture
👤
User Browser
pazlabs.io
→
🔥
Firebase Hosting
pazlabs.io
→
☁️
Cloudflare Tunnel
api.pazlabs.io
↓
⚡
edge-router
Jetson :8000 · FastAPI
↙
↘
🟠
Local Ollama
gemma2 · llama3.2
Confidence > 0.7 → return
☁️
Cloud LLMs
Routed by skill
🟣 Claude — coding🔵 Gemini — math/data🟡 Grok — news🟢 OpenAI — general
Open Source
Repositories
edge-router
FastAPI skill router on Jetson. Runs Ollama locally and escalates to Claude, Gemini, Grok, or OpenAI based on confidence scoring.