A Generalized RAG-LLM Workflow Support Tool for JuliaHealth

This project addresses the challenge of unreliable LLM outputs in complex health informatics workflows by building a domain-grounded Retrieval-Augmented Generation (RAG) system within JuliaHealth. It enables users to convert natural language queries into accurate, executable FunSQL.jl workflows by grounding responses in curated resources like OMOP and OHDSI. The project will deliver a modular baseline RAG pipeline, a reproducible evaluation framework to benchmark query quality, and clear documentation with an end-to-end demonstration for usability and future extension.