Decentralized AI at the Edge of Tomorrow
TANGNET is an ambitious project to create a fully autonomous AI infrastructure using 100 Raspberry Pi nodes and 10 high-end GPU rigs. No cloud dependencies. Complete control. Infinite possibilities.
Run powerful AI models without relying on external APIs or cloud services. Your data stays yours.
Leverage the power of 100 coordinated nodes for parallel processing and fault tolerance.
Start with one Pi, scale to 100. Add GPU nodes as needed. Grow at your own pace.
Run models directly on Raspberry Pi hardware with optimized performance using llama.cpp.
Tailscale-powered VPN mesh for secure, seamless communication between all nodes.
WebSocket-based chat interface with streaming responses and multiple AI personalities.
Hybrid architecture combining edge nodes with powerful GPU rigs for heavy workloads.
Future support for voice transcription, image analysis, and document retrieval.
Every component is open source. Build, modify, and extend as you see fit.
Raspberry Pi 5 running TinyLlama 1.1B with custom chat interface.
2 active nodes: Pi 5 8GB (TinyLlama 1.1B) + Pi 5 16GB (Llama 2 7B). Next: 3 more nodes + RTX 3080.
10 nodes with distributed vector database, persistent memory, and API gateway.
25 nodes with zone-based architecture, redundancy, and voice commands.
100 Pi nodes + 10 RTX 5090 rigs. Complete offline AI ecosystem.
Ready to build your own decentralized AI cluster? Dive into our comprehensive documentation.
Setup Guide View on GitHub 🔮 Live Network Map Architecture Mesh Inventory Llama 2 7B GuideFrom concept to reality - building the hardware that powers TANGNET.
🔬 View Lab Photos