Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SafeMate: A Modular RAG-Based Agent for Context-Aware Emergency Guidance

Junfeng Jiao, Jihyung Park, Yiming Xu, Kristen Sussman, Lucy Atkinson

TL;DR

SafeMate tackles the challenge of making emergency guidance accessible to the general public by combining a Model Context Protocol driven modular agent with retrieval-augmented generation and RAPTOR-based hierarchical retrieval. It grounds responses in authoritative sources from FEMA, CDC, and SOHA, and employs a low-hallucination LLM (o3-mini) with a double verification pass to improve factual reliability. The system architecture supports dynamic routing to external tools, multi-source retrieval, and multi-modal outputs, enabling preparedness planning and real-time crisis guidance. Experimental evaluation on a 100-question benchmark plus a chemical-spill case study demonstrates improved correctness, groundedness, and actionability compared with baselines, highlighting SafeMate’s potential to enhance public safety communication and response in diverse emergency contexts.

Abstract

Despite the abundance of public safety documents and emergency protocols, most individuals remain ill-equipped to interpret and act on such information during crises. Traditional emergency decision support systems (EDSS) are designed for professionals and rely heavily on static documents like PDFs or SOPs, which are difficult for non-experts to navigate under stress. This gap between institutional knowledge and public accessibility poses a critical barrier to effective emergency preparedness and response. We introduce SafeMate, a retrieval-augmented AI assistant that delivers accurate, context-aware guidance to general users in both preparedness and active emergency scenarios. Built on the Model Context Protocol (MCP), SafeMate dynamically routes user queries to tools for document retrieval, checklist generation, and structured summarization. It uses FAISS with cosine similarity to identify relevant content from trusted sources.

SafeMate: A Modular RAG-Based Agent for Context-Aware Emergency Guidance

TL;DR

SafeMate tackles the challenge of making emergency guidance accessible to the general public by combining a Model Context Protocol driven modular agent with retrieval-augmented generation and RAPTOR-based hierarchical retrieval. It grounds responses in authoritative sources from FEMA, CDC, and SOHA, and employs a low-hallucination LLM (o3-mini) with a double verification pass to improve factual reliability. The system architecture supports dynamic routing to external tools, multi-source retrieval, and multi-modal outputs, enabling preparedness planning and real-time crisis guidance. Experimental evaluation on a 100-question benchmark plus a chemical-spill case study demonstrates improved correctness, groundedness, and actionability compared with baselines, highlighting SafeMate’s potential to enhance public safety communication and response in diverse emergency contexts.

Abstract

Despite the abundance of public safety documents and emergency protocols, most individuals remain ill-equipped to interpret and act on such information during crises. Traditional emergency decision support systems (EDSS) are designed for professionals and rely heavily on static documents like PDFs or SOPs, which are difficult for non-experts to navigate under stress. This gap between institutional knowledge and public accessibility poses a critical barrier to effective emergency preparedness and response. We introduce SafeMate, a retrieval-augmented AI assistant that delivers accurate, context-aware guidance to general users in both preparedness and active emergency scenarios. Built on the Model Context Protocol (MCP), SafeMate dynamically routes user queries to tools for document retrieval, checklist generation, and structured summarization. It uses FAISS with cosine similarity to identify relevant content from trusted sources.
Paper Structure (30 sections, 6 equations, 3 figures, 2 tables)

This paper contains 30 sections, 6 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Limitations of Traditional Emergency Support Systems
  4. Retrieval-Augmented Generation and Reasoning with LLMs
  5. E-KELL and Structured Knowledge Reasoning in Emergencies
  6. Cross-Domain Applications of Large Language Models
  7. Methodology
  8. System Architecture Overview
  9. Model Context Protocol (MCP)
  10. Low-Hallucination Language Model: o3-mini
  11. Vector Similarity Search with FAISS
  12. Retrieval-Augmented Generation (RAG) with Authoritative Sources
  13. Hierarchical Retrieval with RAPTOR
  14. Text Chunking and Embedding
  15. Dimensionality Reduction and Clustering
  16. ...and 15 more sections

Figures (3)

  • Figure 1: Overview of the SafeMate architecture.
  • Figure 2: RAPTOR processing detail.
  • Figure 3: GPT as an evaluator.