local-llm-setup

Primary Use Case: Mode B Task Delegation

Mode B is the fast path. run_agent.py sends a lean prompt directly to llama-server — no proxy overhead, no 29K system prompt. Measured: ~2s wall clock for a typical bounded task.