Summary
Reports early experience building an AI-assisted, agent-based ecosystem—RAG grounded in PETSc manual pages and curated documentation—to help users select solvers, debug configurations, and validate results in PETSc.
Keywords
PETScLLM agentsretrieval-augmented generationscientific software usabilitynumerical solversAI for HPC
Cite
@article{2026_smith_petsc_agent_workflows,
title = {Improving Usability and Productivity of PETSc with Agent-Based Workflows},
author = {Barry Smith and Hong Zhang and Junchao Zhang and Satish Balay and Le Chen and Murat Keçeli and Lois Curfman McInnes},
journal = {2026 (preprint/report)},
year = {2026}
}