Research note / structural placeholder
LLM/RAG for scientific models
Placeholder page for agentic interfaces over scientific modeling code and documentation.
Retrieval layer
The retrieval system should index documentation, source files, configuration examples, and prior run logs with provenance preserved.
Agentic workflow
A modeling assistant can propose code changes, prepare experiment directories, run commands, and summarize diagnostics, while write actions remain traceable.
Evaluation
Evaluation should test citation correctness, command safety, configuration fidelity, and whether generated explanations map to actual model code.