Lab: a hardened production endpoint
Put an authenticated, TLS-terminated, rate-limited and monitored llama.cpp server behind Caddy with systemd, then prove it works with verification tests and failure drills.
45 min·Verified 2026-11Hands-on labPro+100 XP
Goal: expose a local model to your team safely, with TLS, per-client API keys, request limits, automatic restarts, monitoring and a firewall, and prove each property with a test rather than assuming it.
You need: a Linux host with an NVIDIA GPU (the same steps work on CPU), a domain name pointing at it, root access, and llama.cpp built or installed (from source). Time: about 45 minutes.
Pro content
Keep reading with Inference Academy Pro
Unlock this full lab, plus every advanced lesson, hands-on lab, the larger playground models and completion certificates.