Deploying Frontier Agentic Technology in MOOSEnger, a Multiphysics-Capable AI Assistant
This work extends MOOSEnger with a harness focused on locally-hosted models that retrieves contextual knowledge from the MOOSE repository, validates and diagnoses the resulting input through interaction with the simulation executable environment, and extracts and stores lessons in a persistent memory.