Docs that teach by running, not reading.
Self-contained resources on inference infrastructure — interactive tutorials and hands-on runbooks. Everything works offline, no tracking, zero prior knowledge assumed.
SMG — Shepherd Model Gateway
SMG Mastery
What a model gateway is and how SMG works inside — from first principles to production patterns. Includes a cache-aware routing simulator, per-chapter quizzes, a glossary, and a cheat sheet. Progress saves in your browser.
open tutorial →SMG Local Lab
Build SMG in Docker and run it against a local Ollama model, or point it at a live engine in a Kubernetes cluster via port-forward. Ships with a battle-tested troubleshooting table for every pitfall we hit.
Ollama Health Shim
A tiny proxy that makes Ollama registrable as an SMG worker — answers /health and /server_info itself and transparently proxies everything else, including streaming. For when SMG logs show Router ready |workers: [].
This is learning material about SMG, built from public sources. The gateway itself lives at smg-project/smg (Apache-2.0). Share freely.