Agentic AI pipeline delivers evidence-grounded neighborhood environmental health answers

[medRxiv] A preprint by Ingebretsen et al. reports the development and evaluation of an agentic AI pipeline that gives evidence-grounded, neighborhood-scale answers to lay persons' health-related environmental questions. Built with LangGraph and OpenAI large language models, the pipeline interprets natural-language queries, selects and invokes external tools, and integrates weather forecasts, air-quality observations, geospatial services and a retrieval-augmented generation module built from asthma-related clinical guidance, with fallback mechanisms when external services are unavailable. The authors evaluated it on a curated benchmark of representative user questions, assessing tool-needed recognition, correct tool selection, avoidance of irrelevant tools, parameter mapping, analytic pathway selection and incorporation of tool outputs. According to the authors, the system integrated heterogeneous environmental and health sources into a unified conversational interface for the Salt Lake Valley, with generally strong tool recognition, parameter mapping and incorporation of retrieved information; they said the modular architecture can accommodate additional data streams, health knowledge sources and regions. The work is a preprint and has not been peer reviewed.