Building Advanced RAG Q&A with Multiple Data Sources Using Langchain: A Multi-Search Agent RAG Application in Ubiquitous Learning

The integration of Large Language Models (LLMs) in Question-Answering (QA) systems has made significant progress, yet they often fail to generate precise answers for queries beyond their training data and hallucinating. To address this, our study develops an advanced Retrieval-Augmented Generation (RAG) pipeline method using the LangChain framework, featuring a decision-making agent that dynamically selects the most effective tools and data sources for accurate and contextually relevant responses. By incorporating multiple data sources and diverse tools, our system mitigates the limitations of traditional LLMs, enhancing their effectiveness in ubiquitous learning environments. Evaluation through various case studies shows significant improvements in accuracy, relevance, and contextual appropriateness. The multi-search agent RAG system efficiently retrieves and synthesizes information, supporting continuous and context-aware learning and enriching the user experience. This research advances AI-based educational technologies, delivering a powerful solution for information retrieval and synthesis, as well as laying the foundations for intelligent question-and-answer systems of the future.