🛠️ Just published: lightchain-worker-toolkit

Announcements
1

PowerShell + Bash automation for the full worker onboarding flow on mainnet & testnet. Built it after burning an afternoon hitting every undocumented gotcha — figured the community shouldn’t have to repeat that.

:link: GitHub - lightchain-protocol/lightchain-worker-toolkit: Batteries-included PowerShell + Bash automation for running a Lightchain mainnet/testnet worker on your own hardware. 9 onboarding phases, 16 named failure modes solved, cross-platform from day one. · GitHub

What’s in it
• 17 PowerShell + 17 Bash scripts — Windows / Linux / macOS native, no WSL needed
• 9 deep-dive docs (architecture, per-OS install, onboarding, day-2 ops, security, FAQ)
• docs/troubleshooting.md documents 16 named failure modes with exact symptoms + fixes
• systemd unit + docker-compose examples
• GitHub Actions linting on every PR

Two real bugs it defaults around for you

  1. SUPPORTED_MODELS=llama3-8b:latest makes every routed job fail silently at stage 5. Must be bare llama3-8b because the worker computes keccak256(name) locally and matches against the on-chain hash. The :latest form gives a different hash and the lookup misses. The toolkit ships with the right value.
  2. On Docker Desktop for Windows, host.docker.internal resolves IPv6-first and stalls the worker’s Go HTTP client to Ollama. The toolkit pins to the explicit IPv4 192.168.65.254:11434.

If you’ve ever stared at worker_ollama_up: 0 for an hour, watched stage 5 fail with model id 0xf4a414fa... and wondered what that hash even is, or hit a 403 trying to gsutil cp chain-node artifacts you didn’t want anyway — there’s a page in the docs for each of those.

MIT licensed. Issues and PRs welcome :saluting_face: