clarus-llm-setup

name	clarus-llm-setup
description	Start the local LLM server for the clarus explain step. Use when Dispatch Brief generation fails or when the explain stage is needed in the demo.
license	Apache-2.0
compatibility	Requires llama.cpp installed (brew install llama.cpp on macOS)
metadata	{"repo":"clarus"}

cd scripts && ./run_llama.sh

Starts llama-server at http://localhost:8080. Default model: bartowski/Llama-3.2-3B-Instruct-GGUF

Then pass --llm-url http://localhost:8080 to eds explain run.