Local-Llm

Experimental: local-first inference governed by explicit policy, measured capability, and strict tool authority.

Concluded prototype: worked on a narrow raw-output path; later evidence showed modern parsed endpoints usually do not need it.