Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{197123,
        author = {D Pavani Sesharatnam and B.S. Sai Ram and T. Vijay Paul and D. Raju and U. Karthik and Dr. Y. Venkat},
        title = {EXPLAINABLE AI CHATBOT USING VECTOR SIMILARITY SEARCH AND LLM'S},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {6141-6147},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=197123},
        abstract = {Large Language Models (LLMs) have achieved unprecedented success in natural language understanding tasks but are plagued by 'black box' architectures that result in hallucinations and a severe lack of data provenance. Retrieval-Augmented Generation (RAG) was proposed to address this issue in LLMs by grounding their outputs on external data. However, in developing the underlying architectures, there was a severe lack of focus on local data privacy and user-centric explainability. This paper proposes a Glass-Box Explainable AI (XAI) framework for a fully offline chatbot that employs high-speed vector similarity search using Facebook AI Similarity Search (FAISS) and local inference using Llama 3. This system bridges the Interpretability Gap through a multi-layered explainability module that provides explicit source attribution, similarity-based confidence scores, and a hard rejection mechanism that completely removes hallucinations at an architectural level. By employing Ollama for local model hosting, this system also solves the Privacy Gap inherent in traditional RAG implementations in the cloud. Empirical evaluation on 60 domain-specific Q&A pairs using the RAGAS framework shows faithfulness of 0.94 and hallucination resistance of 0.96—increases of 74% and 128% over a standalone LLM baseline—thus establishing a new standard in transparent domain-specific AI assistants.},
        keywords = {Explainable AI (XAI), Retrieval-Augmented Generation (RAG), FAISS, Llama 3, Vector Similarity Search, Data Privacy, Local LLMs, Ollama, RAGAS Framework, Confidence Scoring.},
        month = {April},
        }