International Journal of Scientific Research and Engineering Development

Enterprise data has taken off at a high rate in various formats and this has posed a big challenge in effective knowledge retrieval in organizations. Even though such a tool as ChatGPT can adequately respond to general questions, it is not provided with access to organizational data, including internal documents, policies, and reports. This makes organizations need intelligent systems that are able to come up with responses that are solely based on the verified and confidential information of the organizations. In order to overcome these shortcomings, Retrieval-Augmented Generation (RAG) uses information retrieval methods alongside large language models to generate correct and context-relevant responses using enterprise knowledge. Such technologies as FAISS and Pinecone make this process even more efficient as they provide an opportunity to search the large collections of documents by means of semantic searches. The study offers a framework Advanced Multi-Source RAG framework in an enterprise-specific form. The system is built to combine various sources of data such as the PDF documents, the CSV files and the web information with the help of an interactive interface that is created using the Streamlit. It uses the LangChain and LlamaIndex to process, index and the retrieved documents and allows generation of accurate, context aware and reliable responses to any given body of knowledge held privately by an enterprise. The suggested system enhances restoring accuracy and minimizes hallucination as well as decision-making efficiency.

Shifa Bilal Tamboli, Simeen Phiroj Mulani, Arman Tajuddin Shiakh,"Advanced Multi-Source RAG for Enterprise Knowledge Base" International Journal of Scientific Research and Engineering Development, V9(2): Page(2476-2483) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.