🧪 Tangnet Local LLM Node Guide
🖥️ System Overview
- Raspberry Pi 5 (8GB) — Running TinyLlama LLM
- Network Address:
192.168.1.31
- Laptop: RTX 4060, 32GB RAM, Windows
- Main Rig: RTX 3070, 64GB RAM, Windows
🧠 LLM Runtime (TinyLlama)
Run the model:
~/tangnet/llama.cpp/build/bin/llama-run ~/tangnet/llama.cpp/tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf "Your prompt here"
Create a bash alias:
echo "alias tangnet='~/tangnet/llama.cpp/build/bin/llama-run ~/tangnet/llama.cpp/tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf'" >> ~/.bashrc && source ~/.bashrc
Then just run:
tangnet "What's the mission?"
🔐 Connecting to Pi
From laptop or desktop:
ssh brand@192.168.1.31
Transfer a file from laptop to Pi:
scp path/to/your/model.gguf brand@192.168.1.31:~/tangnet/llama.cpp/
Use VNC Viewer (if enabled):
vncviewer 192.168.1.31:1
🔧 Managing the Pi
Start, stop, or reboot properly:
sudo shutdown -h now # Shutdown safely
sudo reboot # Reboot the Pi
Check CPU temperature:
vcgencmd measure_temp
Monitor processes:
htop
Best practice: Shutdown the Pi if not used daily. Leave on for automation or persistent use.
🧪 Handling the Beast
This Pi is now your edge-node for LLM inference. Treat it with respect. Don’t yank the power — always shut down cleanly. For now, you can power it on when needed, SSH/VNC in, and fire up the llama.
Future plans: hook it into a smart outlet and run a local API!
🧠 Rick Note:
"Morty, we got a local brain running on a Pi, and it’s not even melting, Morty. That’s like fitting a black hole in a lunchbox. Don’t screw this up."