agrag-llm-judge

Run headless AgRAG prompts in a shared thread and produce an LLM-as-judge evaluation report. Use for