Skip to content

Latest commit

 

History

32 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Helium — private P2P GPU sharing for people who trust each other

Borrow your friend's idle GPU over an encrypted tunnel, run the workload in an isolated microVM, pay in local credits. No marketplace strangers, no blockchain, no wallet. MIT open source.

Status: technical MVP validated end-to-end (tunnel + microVM + market + auto-matching + TUI). Binary v0.1.0-alpha16: Linux (x86_64/aarch64), macOS, Windows + .deb.

Helium demo

borrower --(WireGuard)-- provider --(tap)--> Firecracker microVM
   |                           |
 CLI / TUI / daemon     SQLite market + credits
   |                           |
 mDNS (LAN) + Kademlia (WAN) discovery

Quick start

Provider (shares RAM/GPU, Linux with KVM):

curl -fsSL https://raw.githubusercontent.com/HeliumXChain/Helium/main/install.sh | sudo bash -s -- --provider --daemon

Borrower (needs compute):

# Linux
curl -fsSL https://raw.githubusercontent.com/HeliumXChain/Helium/main/install.sh | bash -s -- --borrower --endpoint PROVIDER_IP:51820 --pubkey PROVIDER_KEY
# Windows (admin PowerShell)
./install.ps1 -Borrower -Endpoint "PROVIDER_IP:51820" -PeerPubKey "PROVIDER_KEY"

Marketplace (either side):

helium faucet --amount 100                      # demo credits
helium market offer --rtype gpu --amount 32 --price 0.5
helium market request --rtype gpu --amount 16 --max-price 1.0 --hours 2
helium market match --request-id <id> && helium market accept --match-id <id>
helium daemon                                   # auto-matches every 30s
helium dash                                     # fullscreen dashboard

Requirements: Linux 5.10+, /dev/kvm (provider), 8GB RAM recommended.

Layout

helium-mesh/          # the shipped binary (Rust): CLI, daemon, market, tunnel, discovery, VM, TUI
helium/crates/        # original workspace (core matcher, tunnel/VM stubs, CLI)
helium/scripts/       # fc-vm.sh (microVM boot/ssh/expose), fc-provision.sh (guest setup), mlp_train.py (demo)
install.sh            # one-click Linux installer (apt-first, tunnel, firecracker, daemon service)
install.ps1           # Windows installer (binary, PATH, firewall, borrower tunnel)

Validation

Every gate below ran on real hardware (Windows borrower + Ubuntu provider), logged in validation_evidence.md (private):

  • WireGuard tunnel, ping 2–3ms, survives reboots
  • Firecracker boot <1min + SSH, ML training inside the VM (loss 3.27→1.73)
  • 2 manual credit-settled transactions + daemon auto-matching + auto tunnel provisioning
  • mDNS + Kademlia discovery with stable peer IDs, 300MB peer-to-peer transfer hash-verified

License

MIT. Issues and PRs welcome after v0.1 — validation by usage (stars, forks), not interviews.

About

Private P2P GPU sharing for trusted circles — WireGuard tunnels, Firecracker microVMs, local credits. No blockchain. MIT.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Used by

Contributors

Languages