a multi-agent retrieval-augmented generation model with specialized domain agents

Efficiently handling large volumes of customer inquiries across diverse domains is a significant challenge for organizations, especially in data-intensive industries like telecommunications. traditional methods struggle to provide accurate and timely responses as request volumes surge. this paper presents a multi-agent retrieval-augmented generation (rag) system that addresses these challenges by integrating specialized domain agents with an overarching language model supervisor. the system leverages few-shot prompting and the react technique to enhance the supervisor's reasoning and decision-making capabilities. specialized agents employ a novel weighted similarity metric to improve retrieval accuracy for numerical data. the proposed approach was evaluated in a real-world customer service infrastructure, demonstrating superior performance compared to traditional rag systems.