BLOG
Engineering notes.
On governed agents, open-weight models, and self-hosted AI infrastructure.
Anatomy of a governed LLM request
A technical walkthrough of what happens between an agent asking for a completion and a model generating one — JWT auth, OPA policy, rate limits, budgets, semantic cache, routing, and audit.
Introducing Eparch
Why we built an open-source, self-hostable control plane for governed AI agents on open-weight models — and what ships in v1.