This article presents InsuranceRAG, an open-source software system designed to support intelligent interaction with health insurance data using Retrieval-Augmented Generation (RAG). The software provides a Flask-based web interface that enables users to submit natural language queries, upload insurance policy documents, and receive context-aware responses. It follows a modular architecture comprising independent components for conversational assistance, policy recommendation, and document retrieval, all coordinated through a centralized backend. Shared RAG services integrate text embedding models, FAISS-based semantic search, and configurable large language model inference. The open-source implementation promotes reproducibility, extensibility, and practical adoption in insurance-focused natural language processing applications. • Open-source, end-to-end software artifact for insurance RAG applications: Presents InsuranceRAG, a fully implemented and publicly available Retrieval-Augmented Generation system that enables conversational querying, policy recommendation, and clause-level document retrieval from health insurance data. • Modular, reproducible architecture supporting real-world deployment: Implements a Flask-based, multi-module software design with shared embedding, FAISS-powered semantic retrieval, and configurable LLM backends, allowing straightforward extension, replication, and domain adaptation. • Practical impact for research, education, and InsurTech prototyping: Demonstrates a lightweight yet scalable software framework that improves transparency and accessibility of insurance information, serving as a reusable reference system for applied NLP research and early-stage industrial adoption.
Hanmante et al. (Fri,) studied this question.