jingtao-lbl/A2MC-elm: A2MC v2.73

A2MC v2.73 — Agentic Adaptive Multi-target Calibration Autonomous 7-phase calibration framework for E3SM Land Model configurations (ELM-FATES, ELM-ReSOM) combining LLM reasoning, hybrid RAG/GraphRAG knowledge retrieval, and persistent adaptive memory. Highlights Multi-provider AI support (Anthropic Claude, OpenAI GPT, DOE CBorg proxy) Three-level iteration structure (calibration rounds, experiment cycles, skip-testing cycles) Two-tier adaptive memory (generic model knowledge + site-specific calibration lessons) Hybrid RAG (ChromaDB) + GraphRAG (NetworkX) for model documentation retrieval Protocol recommendations in AI diagnosis output Generic/site-specific discovery retrieval pipeline Session-scoped logging and checkpoint/resume across HPC sessions Requirements Python 3.10+ NERSC Perlmutter (or any SLURM-based HPC) API key for Anthropic, OpenAI, or DOE CBorg